US20190065260A1

US20190065260A1 - Technologies for kernel scale-out

Info

Publication number: US20190065260A1
Application number: US15/858,316
Authority: US
Inventors: Susanne M. Balle; Evan Custodio; Francesc Guim Bernat; Slawomir PUTYRSKI
Original assignee: Intel Corp
Current assignee: Intel Corp
Priority date: 2017-08-30
Filing date: 2017-12-29
Publication date: 2019-02-28
Also published as: US20190068509A1; US20190065281A1; CN109426633A; US20190065212A1; DE112018004798T5; US11748172B2; US20190068696A1; US20190065415A1; US11467885B2; US11030017B2; US20190069433A1; CN109428841A; US20190068464A1; CN109428843A; US20190065172A1; WO2019045929A1; US20190068523A1; US20190065261A1; US10888016B2; US20200192710A1

Abstract

Technologies for scaling provisioning of kernel instances in a system as a function of a topology of accelerated kernels include a compute device having a compute engine. The compute engine receives, from a sled, a kernel configuration request to provision a kernel on an accelerator device. The sled is to execute a workload. The kernel accelerates a task in the workload. The compute engine determines, as a function of one or more requirements of the workload, a topology of kernels to service the request. The topology maps data communication between kernels. The compute engine configures the kernel on the accelerator device according to the determined topology. Other embodiments are also described and claimed

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

The present application claims the benefit of Indian Provisional Patent Application No. 201741030632, filed Aug. 30, 2017 and U.S. Provisional Patent Application No. 62/584,401, filed Nov. 10, 2017.

BACKGROUND

In systems that distribute workloads among multiple compute devices (e.g., in a data center), a centralized server may compose nodes of compute devices to process the workloads. Each node represents a logical aggregation of resources (e.g., compute, storage, acceleration, and the like) provided by each compute device. For instance, the node may include a compute device configured with hardware accelerators, such as field-programmable gate array (FPGA) devices and/or graphical processing units (GPUs). Generally, the hardware accelerator improves the execution speed of workload functions. To accelerate a given function of a workload, such as of an application, the centralized server may configure an accelerator device with an accelerated kernel that is suitable for accelerating the task. Once complete, the accelerator device returns data resulting from the accelerated function to the application.

BRIEF DESCRIPTION OF THE DRAWINGS

The concepts described herein are illustrated by way of example and not by way of limitation in the accompanying figures. For simplicity and clarity of illustration, elements illustrated in the figures are not necessarily drawn to scale. Where considered appropriate, reference labels have been repeated among the figures to indicate corresponding or analogous elements.

FIG. 1 is a simplified diagram of at least one embodiment of a data center for executing workloads with disaggregated resources;

FIG. 2 is a simplified diagram of at least one embodiment of a pod of the data center of FIG. 1;

FIG. 3 is a perspective view of at least one embodiment of a rack that may be included in the pod of FIG. 2;

FIG. 4 is a side plan elevation view of the rack of FIG. 3;

FIG. 5 is a perspective view of the rack of FIG. 3 having a sled mounted therein;

FIG. 6 is a is a simplified block diagram of at least one embodiment of a top side of the sled of FIG. 5;

FIG. 7 is a simplified block diagram of at least one embodiment of a bottom side of the sled of FIG. 6;

FIG. 8 is a simplified block diagram of at least one embodiment of a compute sled usable in the data center of FIG. 1;

FIG. 9 is a top perspective view of at least one embodiment of the compute sled of FIG. 8;

FIG. 10 is a simplified block diagram of at least one embodiment of an accelerator sled usable in the data center of FIG. 1;

FIG. 11 is a top perspective view of at least one embodiment of the accelerator sled of FIG. 10;

FIG. 12 is a simplified block diagram of at least one embodiment of a storage sled usable in the data center of FIG. 1;

FIG. 13 is a top perspective view of at least one embodiment of the storage sled of FIG. 12;

FIG. 14 is a simplified block diagram of at least one embodiment of a memory sled usable in the data center of FIG. 1; and

FIG. 15 is a simplified block diagram of a system that may be established within the data center of FIG. 1 to execute workloads with managed nodes composed of disaggregated resources.

FIG. 16 is a simplified block diagram of at least one embodiment of a system for managing a kernel topology for a workload;

FIG. 17 is a simplified block diagram of at least one embodiment of an orchestrator server of the system of FIG. 16;

FIG. 18 is a simplified block diagram of at least one embodiment of an accelerator sled of the system of FIG. 16;

FIG. 19 is a simplified block diagram of at least one embodiment of an environment that may be established by the orchestrator server of FIGS. 16 and 17;

FIG. 20 is a simplified block diagram of at least one embodiment of an environment that may be established by the accelerator sled of FIGS. 16 and 17;

FIG. 21 is a simplified flow diagram of at least one embodiment of a method for determining a topology of accelerated kernels in a system;

FIG. 22 is a simplified flow diagram of at least one embodiment of a method for scaling a configuration of accelerated kernels in a system as a function of a topology of the accelerated kernels;

FIG. 23 is a simplified flow diagram of at least one embodiment of a method for provisioning additional instances of a kernel executing on an accelerator device as a function of a power state associated with the accelerator device;

FIGS. 24A and 24B are diagrams of an example embodiment of provisioning additional instances of a kernel executing on an accelerator device as a function of a power state associated with the accelerator device;

FIG. 25 is a simplified flow diagram of at least one embodiment of a method for provisioning, via an in-band communication by an accelerator device, an accelerated kernel on another accelerator device; and

FIGS. 26A and 26B are diagrams of an example embodiment of provisioning, via an in-band communication by an accelerator device, an accelerated kernel on another accelerator device.

DETAILED DESCRIPTION OF THE DRAWINGS

While the concepts of the present disclosure are susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described herein in detail. It should be understood, however, that there is no intent to limit the concepts of the present disclosure to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives consistent with the present disclosure and the appended claims.
References in the specification to “one embodiment,” “an embodiment,” “an illustrative embodiment,” etc., indicate that the embodiment described may include a particular feature, structure, or characteristic, but every embodiment may or may not necessarily include that particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. Additionally, it should be appreciated that items included in a list in the form of “at least one A, B, and C” can mean (A); (B); (C); (A and B); (A and C); (B and C); or (A, B, and C). Similarly, items listed in the form of “at least one of A, B, or C” can mean (A); (B); (C); (A and B); (A and C); (B and C); or (A, B, and C).
The disclosed embodiments may be implemented, in some cases, in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried by or stored on a transitory or non-transitory machine-readable (e.g., computer-readable) storage medium, which may be read and executed by one or more processors. A machine-readable storage medium may be embodied as any storage device, mechanism, or other physical structure for storing or transmitting information in a form readable by a machine (e.g., a volatile or non-volatile memory, a media disc, or other media device).
In the drawings, some structural or method features may be shown in specific arrangements and/or orderings. However, it should be appreciated that such specific arrangements and/or orderings may not be required. Rather, in some embodiments, such features may be arranged in a different manner and/or order than shown in the illustrative figures. Additionally, the inclusion of a structural or method feature in a particular figure is not meant to imply that such feature is required in all embodiments and, in some embodiments, may not be included or may be combined with other features.
Referring now to FIG. 1, a data center 100 in which disaggregated resources may cooperatively execute one or more workloads (e.g., applications on behalf of customers) includes multiple pods 110, 120, 130, 140, each of which includes one or more rows of racks. As described in more detail herein, each rack houses multiple sleds, which each may be embodied as a compute device, such as a server, that is primarily equipped with a particular type of resource (e.g., memory devices, data storage devices, accelerator devices, general purpose processors). In the illustrative embodiment, the sleds in each pod 110, 120, 130, 140 are connected to multiple pod switches (e.g., switches that route data communications to and from sleds within the pod). The pod switches, in turn, connect with spine switches 150 that switch communications among pods (e.g., the pods 110, 120, 130, 140) in the data center 100. In some embodiments, the sleds may be connected with a fabric using Intel Omni-Path technology. As described in more detail herein, resources within sleds in the data center 100 may be allocated to a group (referred to herein as a “managed node”) containing resources from one or more other sleds to be collectively utilized in the execution of a workload. The workload can execute as if the resources belonging to the managed node were located on the same sled. The resources in a managed node may even belong to sleds belonging to different racks, and even to different pods 110, 120, 130, 140. Some resources of a single sled may be allocated to one managed node while other resources of the same sled are allocated to a different managed node (e.g., one processor assigned to one managed node and another processor of the same sled assigned to a different managed node). By disaggregating resources to sleds comprised predominantly of a single type of resource (e.g., compute sleds comprising primarily compute resources, memory sleds containing primarily memory resources), and selectively allocating and deallocating the disaggregated resources to form a managed node assigned to execute a workload, the data center 100 provides more efficient resource usage over typical data centers comprised of hyperconverged servers containing compute, memory, storage and perhaps additional resources). As such, the data center 100 may provide greater performance (e.g., throughput, operations per second, latency, etc.) than a typical data center that has the same number of resources.
Referring now to FIG. 2, the pod 110, in the illustrative embodiment, includes a set of rows 200, 210, 220, 230 of racks 240. Each rack 240 may house multiple sleds (e.g., sixteen sleds) and provide power and data connections to the housed sleds, as described in more detail herein. In the illustrative embodiment, the racks in each row 200, 210, 220, 230 are connected to multiple pod switches 250, 260. The pod switch 250 includes a set of ports 252 to which the sleds of the racks of the pod 110 are connected and another set of ports 254 that connect the pod 110 to the spine switches 150 to provide connectivity to other pods in the data center 100. Similarly, the pod switch 260 includes a set of ports 262 to which the sleds of the racks of the pod 110 are connected and a set of ports 264 that connect the pod 110 to the spine switches 150. As such, the use of the pair of switches 250, 260 provides an amount of redundancy to the pod 110. For example, if either of the switches 250, 260 fails, the sleds in the pod 110 may still maintain data communication with the remainder of the data center 100 (e.g., sleds of other pods) through the other switch 250, 260. Furthermore, in the illustrative embodiment, the switches 150, 250, 260 may be embodied as dual-mode optical switches, capable of routing both Ethernet protocol communications carrying Internet Protocol (IP) packets and communications according to a second, high-performance link-layer protocol (e.g., Intel's Omni-Path Architecture's, Infiniband) via optical signaling media of an optical fabric.
It should be appreciated that each of the other pods 120, 130, 140 (as well as any additional pods of the data center 100) may be similarly structured as, and have components similar to, the pod 110 shown in and described in regard to FIG. 2 (e.g., each pod may have rows of racks housing multiple sleds as described above). Additionally, while two pod switches 250, 260 are shown, it should be understood that in other embodiments, each pod 110, 120, 130, 140 may be connected to different number of pod switches (e.g., providing even more failover capacity).
Referring now to FIGS. 3-5, each illustrative rack 240 of the data center 100 includes two elongated support posts 302, 304, which are arranged vertically. For example, the elongated support posts 302, 304 may extend upwardly from a floor of the data center 100 when deployed. The rack 240 also includes one or more horizontal pairs 310 of elongated support arms 312 (identified in FIG. 3 via a dashed ellipse) configured to support a sled of the data center 100 as discussed below. One elongated support arm 312 of the pair of elongated support arms 312 extends outwardly from the elongated support post 302 and the other elongated support arm 312 extends outwardly from the elongated support post 304.
In the illustrative embodiments, each sled of the data center 100 is embodied as a chassis-less sled. That is, each sled has a chassis-less circuit board substrate on which physical resources (e.g., processors, memory, accelerators, storage, etc.) are mounted as discussed in more detail below. As such, the rack 240 is configured to receive the chassis-less sleds. For example, each pair 310 of elongated support arms 312 defines a sled slot 320 of the rack 240, which is configured to receive a corresponding chassis-less sled. To do so, each illustrative elongated support arm 312 includes a circuit board guide 330 configured to receive the chassis-less circuit board substrate of the sled. Each circuit board guide 330 is secured to, or otherwise mounted to, a top side 332 of the corresponding elongated support arm 312. For example, in the illustrative embodiment, each circuit board guide 330 is mounted at a distal end of the corresponding elongated support arm 312 relative to the corresponding elongated support post 302, 304. For clarity of the Figures, not every circuit board guide 330 may be referenced in each Figure.
Each circuit board guide 330 includes an inner wall that defines a circuit board slot 380 configured to receive the chassis-less circuit board substrate of a sled 400 when the sled 400 is received in the corresponding sled slot 320 of the rack 240. To do so, as shown in FIG. 4, a user (or robot) aligns the chassis-less circuit board substrate of an illustrative chassis-less sled 400 to a sled slot 320. The user, or robot, may then slide the chassis-less circuit board substrate forward into the sled slot 320 such that each side edge 414 of the chassis-less circuit board substrate is received in a corresponding circuit board slot 380 of the circuit board guides 330 of the pair 310 of elongated support arms 312 that define the corresponding sled slot 320 as shown in FIG. 4. By having robotically accessible and robotically manipulable sleds comprising disaggregated resources, each type of resource can be upgraded independently of each other and at their own optimized refresh rate. Furthermore, the sleds are configured to blindly mate with power and data communication cables in each rack 240, enhancing their ability to be quickly removed, upgraded, reinstalled, and/or replaced. As such, in some embodiments, the data center 100 may operate (e.g., execute workloads, undergo maintenance and/or upgrades, etc.) without human involvement on the data center floor. In other embodiments, a human may facilitate one or more maintenance or upgrade operations in the data center 100.
It should be appreciated that each circuit board guide 330 is dual sided. That is, each circuit board guide 330 includes an inner wall that defines a circuit board slot 380 on each side of the circuit board guide 330. In this way, each circuit board guide 330 can support a chassis-less circuit board substrate on either side. As such, a single additional elongated support post may be added to the rack 240 to turn the rack 240 into a two-rack solution that can hold twice as many sled slots 320 as shown in FIG. 3. The illustrative rack 240 includes seven pairs 310 of elongated support arms 312 that define a corresponding seven sled slots 320, each configured to receive and support a corresponding sled 400 as discussed above. Of course, in other embodiments, the rack 240 may include additional or fewer pairs 310 of elongated support arms 312 (i.e., additional or fewer sled slots 320). It should be appreciated that because the sled 400 is chassis-less, the sled 400 may have an overall height that is different than typical servers. As such, in some embodiments, the height of each sled slot 320 may be shorter than the height of a typical server (e.g., shorter than a single rank unit, “1 U”). That is, the vertical distance between each pair 310 of elongated support arms 312 may be less than a standard rack unit “1 U.” Additionally, due to the relative decrease in height of the sled slots 320, the overall height of the rack 240 in some embodiments may be shorter than the height of traditional rack enclosures. For example, in some embodiments, each of the elongated support posts 302, 304 may have a length of six feet or less. Again, in other embodiments, the rack 240 may have different dimensions. Further, it should be appreciated that the rack 240 does not include any walls, enclosures, or the like. Rather, the rack 240 is an enclosure-less rack that is opened to the local environment. Of course, in some cases, an end plate may be attached to one of the elongated support posts 302, 304 in those situations in which the rack 240 forms an end-of-row rack in the data center 100.
In some embodiments, various interconnects may be routed upwardly or downwardly through the elongated support posts 302, 304. To facilitate such routing, each elongated support post 302, 304 includes an inner wall that defines an inner chamber in which the interconnect may be located. The interconnects routed through the elongated support posts 302, 304 may be embodied as any type of interconnects including, but not limited to, data or communication interconnects to provide communication connections to each sled slot 320, power interconnects to provide power to each sled slot 320, and/or other types of interconnects.
The rack 240, in the illustrative embodiment, includes a support platform on which a corresponding optical data connector (not shown) is mounted. Each optical data connector is associated with a corresponding sled slot 320 and is configured to mate with an optical data connector of a corresponding sled 400 when the sled 400 is received in the corresponding sled slot 320. In some embodiments, optical connections between components (e.g., sleds, racks, and switches) in the data center 100 are made with a blind mate optical connection. For example, a door on each cable may prevent dust from contaminating the fiber inside the cable. In the process of connecting to a blind mate optical connector mechanism, the door is pushed open when the end of the cable enters the connector mechanism. Subsequently, the optical fiber inside the cable enters a gel within the connector mechanism and the optical fiber of one cable comes into contact with the optical fiber of another cable within the gel inside the connector mechanism.
The illustrative rack 240 also includes a fan array 370 coupled to the cross-support arms of the rack 240. The fan array 370 includes one or more rows of cooling fans 372, which are aligned in a horizontal line between the elongated support posts 302, 304. In the illustrative embodiment, the fan array 370 includes a row of cooling fans 372 for each sled slot 320 of the rack 240. As discussed above, each sled 400 does not include any on-board cooling system in the illustrative embodiment and, as such, the fan array 370 provides cooling for each sled 400 received in the rack 240. Each rack 240, in the illustrative embodiment, also includes a power supply associated with each sled slot 320. Each power supply is secured to one of the elongated support arms 312 of the pair 310 of elongated support arms 312 that define the corresponding sled slot 320. For example, the rack 240 may include a power supply coupled or secured to each elongated support arm 312 extending from the elongated support post 302. Each power supply includes a power connector configured to mate with a power connector of the sled 400 when the sled 400 is received in the corresponding sled slot 320. In the illustrative embodiment, the sled 400 does not include any on-board power supply and, as such, the power supplies provided in the rack 240 supply power to corresponding sleds 400 when mounted to the rack 240.
Referring now to FIG. 6, the sled 400, in the illustrative embodiment, is configured to be mounted in a corresponding rack 240 of the data center 100 as discussed above. In some embodiments, each sled 400 may be optimized or otherwise configured for performing particular tasks, such as compute tasks, acceleration tasks, data storage tasks, etc. For example, the sled 400 may be embodied as a compute sled 800 as discussed below in regard to FIGS. 8-9, an accelerator sled 1000 as discussed below in regard to FIGS. 10-11, a storage sled 1200 as discussed below in regard to FIGS. 12-13, or as a sled optimized or otherwise configured to perform other specialized tasks, such as a memory sled 1400, discussed below in regard to FIG. 14.
As discussed above, the illustrative sled 400 includes a chassis-less circuit board substrate 602, which supports various physical resources (e.g., electrical components) mounted thereon. It should be appreciated that the circuit board substrate 602 is “chassis-less” in that the sled 400 does not include a housing or enclosure. Rather, the chassis-less circuit board substrate 602 is open to the local environment. The chassis-less circuit board substrate 602 may be formed from any material capable of supporting the various electrical components mounted thereon. For example, in an illustrative embodiment, the chassis-less circuit board substrate 602 is formed from an FR-4 glass-reinforced epoxy laminate material. Of course, other materials may be used to form the chassis-less circuit board substrate 602 in other embodiments.
As discussed in more detail below, the chassis-less circuit board substrate 602 includes multiple features that improve the thermal cooling characteristics of the various electrical components mounted on the chassis-less circuit board substrate 602. As discussed, the chassis-less circuit board substrate 602 does not include a housing or enclosure, which may improve the airflow over the electrical components of the sled 400 by reducing those structures that may inhibit air flow. For example, because the chassis-less circuit board substrate 602 is not positioned in an individual housing or enclosure, there is no backplane (e.g., a backplate of the chassis) to the chassis-less circuit board substrate 602, which could inhibit air flow across the electrical components. Additionally, the chassis-less circuit board substrate 602 has a geometric shape configured to reduce the length of the airflow path across the electrical components mounted to the chassis-less circuit board substrate 602. For example, the illustrative chassis-less circuit board substrate 602 has a width 604 that is greater than a depth 606 of the chassis-less circuit board substrate 602. In one particular embodiment, for example, the chassis-less circuit board substrate 602 has a width of about 21 inches and a depth of about 9 inches, compared to a typical server that has a width of about 17 inches and a depth of about 39 inches. As such, an airflow path 608 that extends from a front edge 610 of the chassis-less circuit board substrate 602 toward a rear edge 612 has a shorter distance relative to typical servers, which may improve the thermal cooling characteristics of the sled 400. Furthermore, although not illustrated in FIG. 6, the various physical resources mounted to the chassis-less circuit board substrate 602 are mounted in corresponding locations such that no two substantively heat-producing electrical components shadow each other as discussed in more detail below. That is, no two electrical components, which produce appreciable heat during operation (i.e., greater than a nominal heat sufficient enough to adversely impact the cooling of another electrical component), are mounted to the chassis-less circuit board substrate 602 linearly in-line with each other along the direction of the airflow path 608 (i.e., along a direction extending from the front edge 610 toward the rear edge 612 of the chassis-less circuit board substrate 602).
As discussed above, the illustrative sled 400 includes one or more physical resources 620 mounted to a top side 650 of the chassis-less circuit board substrate 602. Although two physical resources 620 are shown in FIG. 6, it should be appreciated that the sled 400 may include one, two, or more physical resources 620 in other embodiments. The physical resources 620 may be embodied as any type of processor, controller, or other compute circuit capable of performing various tasks such as compute functions and/or controlling the functions of the sled 400 depending on, for example, the type or intended functionality of the sled 400. For example, as discussed in more detail below, the physical resources 620 may be embodied as high-performance processors in embodiments in which the sled 400 is embodied as a compute sled, as accelerator co-processors or circuits in embodiments in which the sled 400 is embodied as an accelerator sled, storage controllers in embodiments in which the sled 400 is embodied as a storage sled, or a set of memory devices in embodiments in which the sled 400 is embodied as a memory sled.
The sled 400 also includes one or more additional physical resources 630 mounted to the top side 650 of the chassis-less circuit board substrate 602. In the illustrative embodiment, the additional physical resources include a network interface controller (NIC) as discussed in more detail below. Of course, depending on the type and functionality of the sled 400, the physical resources 630 may include additional or other electrical components, circuits, and/or devices in other embodiments.
The physical resources 620 are communicatively coupled to the physical resources 630 via an input/output (I/O) subsystem 622. The I/O subsystem 622 may be embodied as circuitry and/or components to facilitate input/output operations with the physical resources 620, the physical resources 630, and/or other components of the sled 400. For example, the I/O subsystem 622 may be embodied as, or otherwise include, memory controller hubs, input/output control hubs, integrated sensor hubs, firmware devices, communication links (e.g., point-to-point links, bus links, wires, cables, light guides, printed circuit board traces, etc.), and/or other components and subsystems to facilitate the input/output operations. In the illustrative embodiment, the I/O subsystem 622 is embodied as, or otherwise includes, a double data rate 4 (DDR4) data bus or a DDR5 data bus.
In some embodiments, the sled 400 may also include a resource-to-resource interconnect 624. The resource-to-resource interconnect 624 may be embodied as any type of communication interconnect capable of facilitating resource-to-resource communications. In the illustrative embodiment, the resource-to-resource interconnect 624 is embodied as a high-speed point-to-point interconnect (e.g., faster than the I/O subsystem 622). For example, the resource-to-resource interconnect 624 may be embodied as a QuickPath Interconnect (QPI), an UltraPath Interconnect (UPI), or other high-speed point-to-point interconnect dedicated to resource-to-resource communications.
The sled 400 also includes a power connector 640 configured to mate with a corresponding power connector of the rack 240 when the sled 400 is mounted in the corresponding rack 240. The sled 400 receives power from a power supply of the rack 240 via the power connector 640 to supply power to the various electrical components of the sled 400. That is, the sled 400 does not include any local power supply (i.e., an on-board power supply) to provide power to the electrical components of the sled 400. The exclusion of a local or on-board power supply facilitates the reduction in the overall footprint of the chassis-less circuit board substrate 602, which may increase the thermal cooling characteristics of the various electrical components mounted on the chassis-less circuit board substrate 602 as discussed above. In some embodiments, power is provided to the processors 820 through vias directly under the processors 820 (e.g., through the bottom side 750 of the chassis-less circuit board substrate 602), providing an increased thermal budget, additional current and/or voltage, and better voltage control over typical boards.
In some embodiments, the sled 400 may also include mounting features 642 configured to mate with a mounting arm, or other structure, of a robot to facilitate the placement of the sled 600 in a rack 240 by the robot. The mounting features 642 may be embodied as any type of physical structures that allow the robot to grasp the sled 400 without damaging the chassis-less circuit board substrate 602 or the electrical components mounted thereto. For example, in some embodiments, the mounting features 642 may be embodied as non-conductive pads attached to the chassis-less circuit board substrate 602. In other embodiments, the mounting features may be embodied as brackets, braces, or other similar structures attached to the chassis-less circuit board substrate 602. The particular number, shape, size, and/or make-up of the mounting feature 642 may depend on the design of the robot configured to manage the sled 400.
Referring now to FIG. 7, in addition to the physical resources 630 mounted on the top side 650 of the chassis-less circuit board substrate 602, the sled 400 also includes one or more memory devices 720 mounted to a bottom side 750 of the chassis-less circuit board substrate 602. That is, the chassis-less circuit board substrate 602 is embodied as a double-sided circuit board. The physical resources 620 are communicatively coupled to the memory devices 720 via the I/O subsystem 622. For example, the physical resources 620 and the memory devices 720 may be communicatively coupled by one or more vias extending through the chassis-less circuit board substrate 602. Each physical resource 620 may be communicatively coupled to a different set of one or more memory devices 720 in some embodiments. Alternatively, in other embodiments, each physical resource 620 may be communicatively coupled to each memory devices 720.
The memory devices 720 may be embodied as any type of memory device capable of storing data for the physical resources 620 during operation of the sled 400, such as any type of volatile (e.g., dynamic random access memory (DRAM), etc.) or non-volatile memory. Volatile memory may be a storage medium that requires power to maintain the state of data stored by the medium. Non-limiting examples of volatile memory may include various types of random access memory (RAM), such as dynamic random access memory (DRAM) or static random access memory (SRAM). One particular type of DRAM that may be used in a memory module is synchronous dynamic random access memory (SDRAM). In particular embodiments, DRAM of a memory component may comply with a standard promulgated by JEDEC, such as JESD79F for DDR SDRAM, JESD79-2F for DDR2 SDRAM, JESD79-3F for DDR3 SDRAM, JESD79-4A for DDR4 SDRAM, JESD209 for Low Power DDR (LPDDR), JESD209-2 for LPDDR2, JESD209-3 for LPDDR3, and JESD209-4 for LPDDR4 (these standards are available at www.jedec.org). Such standards (and similar standards) may be referred to as DDR-based standards and communication interfaces of the storage devices that implement such standards may be referred to as DDR-based interfaces.
In one embodiment, the memory device is a block addressable memory device, such as those based on NAND or NOR technologies. A memory device may also include next-generation nonvolatile devices, such as Intel 3D XPoint™ memory or other byte addressable write-in-place nonvolatile memory devices. In one embodiment, the memory device may be or may include memory devices that use chalcogenide glass, multi-threshold level NAND flash memory, NOR flash memory, single or multi-level Phase Change Memory (PCM), a resistive memory, nanowire memory, ferroelectric transistor random access memory (FeTRAM), anti-ferroelectric memory, magnetoresistive random access memory (MRAM) memory that incorporates memristor technology, resistive memory including the metal oxide base, the oxygen vacancy base and the conductive bridge Random Access Memory (CB-RAM), or spin transfer torque (STT)-MRAM, a spintronic magnetic junction memory based device, a magnetic tunneling junction (MTJ) based device, a DW (Domain Wall) and SOT (Spin Orbit Transfer) based device, a thyristor based memory device, or a combination of any of the above, or other memory. The memory device may refer to the die itself and/or to a packaged memory product. In some embodiments, the memory device may comprise a transistor-less stackable cross point architecture in which memory cells sit at the intersection of word lines and bit lines and are individually addressable and in which bit storage is based on a change in bulk resistance.
Referring now to FIG. 8, in some embodiments, the sled 400 may be embodied as a compute sled 800. The compute sled 800 is optimized, or otherwise configured, to perform compute tasks. Of course, as discussed above, the compute sled 800 may rely on other sleds, such as acceleration sleds and/or storage sleds, to perform such compute tasks. The compute sled 800 includes various physical resources (e.g., electrical components) similar to the physical resources of the sled 400, which have been identified in FIG. 8 using the same reference numbers. The description of such components provided above in regard to FIGS. 6 and 7 applies to the corresponding components of the compute sled 800 and is not repeated herein for clarity of the description of the compute sled 800.
In the illustrative compute sled 800, the physical resources 620 are embodied as processors 820. Although only two processors 820 are shown in FIG. 8, it should be appreciated that the compute sled 800 may include additional processors 820 in other embodiments. Illustratively, the processors 820 are embodied as high-performance processors 820 and may be configured to operate at a relatively high power rating. Although the processors 820 generate additional heat operating at power ratings greater than typical processors (which operate at around 155-230 W), the enhanced thermal cooling characteristics of the chassis-less circuit board substrate 602 discussed above facilitate the higher power operation. For example, in the illustrative embodiment, the processors 820 are configured to operate at a power rating of at least 250 W. In some embodiments, the processors 820 may be configured to operate at a power rating of at least 350 W.
In some embodiments, the compute sled 800 may also include a processor-to-processor interconnect 842. Similar to the resource-to-resource interconnect 624 of the sled 400 discussed above, the processor-to-processor interconnect 842 may be embodied as any type of communication interconnect capable of facilitating processor-to-processor interconnect 842 communications. In the illustrative embodiment, the processor-to-processor interconnect 842 is embodied as a high-speed point-to-point interconnect (e.g., faster than the I/O subsystem 622). For example, the processor-to-processor interconnect 842 may be embodied as a QuickPath Interconnect (QPI), an UltraPath Interconnect (UPI), or other high-speed point-to-point interconnect dedicated to processor-to-processor communications.
The compute sled 800 also includes a communication circuit 830. The illustrative communication circuit 830 includes a network interface controller (NIC) 832, which may also be referred to as a host fabric interface (HFI). The NIC 832 may be embodied as, or otherwise include, any type of integrated circuit, discrete circuits, controller chips, chipsets, add-in-boards, daughtercards, network interface cards, other devices that may be used by the compute sled 800 to connect with another compute device (e.g., with other sleds 400). In some embodiments, the NIC 832 may be embodied as part of a system-on-a-chip (SoC) that includes one or more processors, or included on a multichip package that also contains one or more processors. In some embodiments, the NIC 832 may include a local processor (not shown) and/or a local memory (not shown) that are both local to the NIC 832. In such embodiments, the local processor of the NIC 832 may be capable of performing one or more of the functions of the processors 820. Additionally or alternatively, in such embodiments, the local memory of the NIC 832 may be integrated into one or more components of the compute sled at the board level, socket level, chip level, and/or other levels.
The communication circuit 830 is communicatively coupled to an optical data connector 834. The optical data connector 834 is configured to mate with a corresponding optical data connector of the rack 240 when the compute sled 800 is mounted in the rack 240. Illustratively, the optical data connector 834 includes a plurality of optical fibers which lead from a mating surface of the optical data connector 834 to an optical transceiver 836. The optical transceiver 836 is configured to convert incoming optical signals from the rack-side optical data connector to electrical signals and to convert electrical signals to outgoing optical signals to the rack-side optical data connector. Although shown as forming part of the optical data connector 834 in the illustrative embodiment, the optical transceiver 836 may form a portion of the communication circuit 830 in other embodiments.
In some embodiments, the compute sled 800 may also include an expansion connector 840. In such embodiments, the expansion connector 840 is configured to mate with a corresponding connector of an expansion chassis-less circuit board substrate to provide additional physical resources to the compute sled 800. The additional physical resources may be used, for example, by the processors 820 during operation of the compute sled 800. The expansion chassis-less circuit board substrate may be substantially similar to the chassis-less circuit board substrate 602 discussed above and may include various electrical components mounted thereto. The particular electrical components mounted to the expansion chassis-less circuit board substrate may depend on the intended functionality of the expansion chassis-less circuit board substrate. For example, the expansion chassis-less circuit board substrate may provide additional compute resources, memory resources, and/or storage resources. As such, the additional physical resources of the expansion chassis-less circuit board substrate may include, but is not limited to, processors, memory devices, storage devices, and/or accelerator circuits including, for example, field programmable gate arrays (FPGA), application-specific integrated circuits (ASICs), security co-processors, graphics processing units (GPUs), machine learning circuits, or other specialized processors, controllers, devices, and/or circuits.
Referring now to FIG. 9, an illustrative embodiment of the compute sled 800 is shown. As shown, the processors 820, communication circuit 830, and optical data connector 834 are mounted to the top side 650 of the chassis-less circuit board substrate 602. Any suitable attachment or mounting technology may be used to mount the physical resources of the compute sled 800 to the chassis-less circuit board substrate 602. For example, the various physical resources may be mounted in corresponding sockets (e.g., a processor socket), holders, or brackets. In some cases, some of the electrical components may be directly mounted to the chassis-less circuit board substrate 602 via soldering or similar techniques.
As discussed above, the individual processors 820 and communication circuit 830 are mounted to the top side 650 of the chassis-less circuit board substrate 602 such that no two heat-producing, electrical components shadow each other. In the illustrative embodiment, the processors 820 and communication circuit 830 are mounted in corresponding locations on the top side 650 of the chassis-less circuit board substrate 602 such that no two of those physical resources are linearly in-line with others along the direction of the airflow path 608. It should be appreciated that, although the optical data connector 834 is in-line with the communication circuit 830, the optical data connector 834 produces no or nominal heat during operation.
The memory devices 720 of the compute sled 800 are mounted to the bottom side 750 of the of the chassis-less circuit board substrate 602 as discussed above in regard to the sled 400. Although mounted to the bottom side 750, the memory devices 720 are communicatively coupled to the processors 820 located on the top side 650 via the I/O subsystem 622. Because the chassis-less circuit board substrate 602 is embodied as a double-sided circuit board, the memory devices 720 and the processors 820 may be communicatively coupled by one or more vias, connectors, or other mechanisms extending through the chassis-less circuit board substrate 602. Of course, each processor 820 may be communicatively coupled to a different set of one or more memory devices 720 in some embodiments. Alternatively, in other embodiments, each processor 820 may be communicatively coupled to each memory device 720. In some embodiments, the memory devices 720 may be mounted to one or more memory mezzanines on the bottom side of the chassis-less circuit board substrate 602 and may interconnect with a corresponding processor 820 through a ball-grid array.
Each of the processors 820 includes a heatsink 850 secured thereto. Due to the mounting of the memory devices 720 to the bottom side 750 of the chassis-less circuit board substrate 602 (as well as the vertical spacing of the sleds 400 in the corresponding rack 240), the top side 650 of the chassis-less circuit board substrate 602 includes additional “free” area or space that facilitates the use of heatsinks 850 having a larger size relative to traditional heatsinks used in typical servers. Additionally, due to the improved thermal cooling characteristics of the chassis-less circuit board substrate 602, none of the processor heatsinks 850 include cooling fans attached thereto. That is, each of the heatsinks 850 is embodied as a fan-less heatsinks.
Referring now to FIG. 10, in some embodiments, the sled 400 may be embodied as an accelerator sled 1000. The accelerator sled 1000 is optimized, or otherwise configured, to perform specialized compute tasks, such as machine learning, encryption, hashing, or other computational-intensive task. In some embodiments, for example, a compute sled 800 may offload tasks to the accelerator sled 1000 during operation. The accelerator sled 1000 includes various components similar to components of the sled 400 and/or compute sled 800, which have been identified in FIG. 10 using the same reference numbers. The description of such components provided above in regard to FIGS. 6, 7, and 8 apply to the corresponding components of the accelerator sled 1000 and is not repeated herein for clarity of the description of the accelerator sled 1000.
In the illustrative accelerator sled 1000, the physical resources 620 are embodied as accelerator circuits 1020. Although only two accelerator circuits 1020 are shown in FIG. 10, it should be appreciated that the accelerator sled 1000 may include additional accelerator circuits 1020 in other embodiments. For example, as shown in FIG. 11, the accelerator sled 1000 may include four accelerator circuits 1020 in some embodiments. The accelerator circuits 1020 may be embodied as any type of processor, co-processor, compute circuit, or other device capable of performing compute or processing operations. For example, the accelerator circuits 1020 may be embodied as, for example, field programmable gate arrays (FPGA), application-specific integrated circuits (ASICs), security co-processors, graphics processing units (GPUs), machine learning circuits, or other specialized processors, controllers, devices, and/or circuits.
In some embodiments, the accelerator sled 1000 may also include an accelerator-to-accelerator interconnect 1042. Similar to the resource-to-resource interconnect 624 of the sled 600 discussed above, the accelerator-to-accelerator interconnect 1042 may be embodied as any type of communication interconnect capable of facilitating accelerator-to-accelerator communications. In the illustrative embodiment, the accelerator-to-accelerator interconnect 1042 is embodied as a high-speed point-to-point interconnect (e.g., faster than the I/O subsystem 622). For example, the accelerator-to-accelerator interconnect 1042 may be embodied as a QuickPath Interconnect (QPI), an UltraPath Interconnect (UPI), or other high-speed point-to-point interconnect dedicated to processor-to-processor communications. In some embodiments, the accelerator circuits 1020 may be daisy-chained with a primary accelerator circuit 1020 connected to the NIC 832 and memory 720 through the I/O subsystem 622 and a secondary accelerator circuit 1020 connected to the NIC 832 and memory 720 through a primary accelerator circuit 1020.
Referring now to FIG. 11, an illustrative embodiment of the accelerator sled 1000 is shown. As discussed above, the accelerator circuits 1020, communication circuit 830, and optical data connector 834 are mounted to the top side 650 of the chassis-less circuit board substrate 602. Again, the individual accelerator circuits 1020 and communication circuit 830 are mounted to the top side 650 of the chassis-less circuit board substrate 602 such that no two heat-producing, electrical components shadow each other as discussed above. The memory devices 720 of the accelerator sled 1000 are mounted to the bottom side 750 of the of the chassis-less circuit board substrate 602 as discussed above in regard to the sled 600. Although mounted to the bottom side 750, the memory devices 720 are communicatively coupled to the accelerator circuits 1020 located on the top side 650 via the I/O subsystem 622 (e.g., through vias). Further, each of the accelerator circuits 1020 may include a heatsink 1070 that is larger than a traditional heatsink used in a server. As discussed above with reference to the heatsinks 870, the heatsinks 1070 may be larger than tradition heatsinks because of the “free” area provided by the memory devices 750 being located on the bottom side 750 of the chassis-less circuit board substrate 602 rather than on the top side 650.
Referring now to FIG. 12, in some embodiments, the sled 400 may be embodied as a storage sled 1200. The storage sled 1200 is optimized, or otherwise configured, to store data in a data storage 1250 local to the storage sled 1200. For example, during operation, a compute sled 800 or an accelerator sled 1000 may store and retrieve data from the data storage 1250 of the storage sled 1200. The storage sled 1200 includes various components similar to components of the sled 400 and/or the compute sled 800, which have been identified in FIG. 12 using the same reference numbers. The description of such components provided above in regard to FIGS. 6, 7, and 8 apply to the corresponding components of the storage sled 1200 and is not repeated herein for clarity of the description of the storage sled 1200.
In the illustrative storage sled 1200, the physical resources 620 are embodied as storage controllers 1220. Although only two storage controllers 1220 are shown in FIG. 12, it should be appreciated that the storage sled 1200 may include additional storage controllers 1220 in other embodiments. The storage controllers 1220 may be embodied as any type of processor, controller, or control circuit capable of controlling the storage and retrieval of data into the data storage 1250 based on requests received via the communication circuit 830. In the illustrative embodiment, the storage controllers 1220 are embodied as relatively low-power processors or controllers. For example, in some embodiments, the storage controllers 1220 may be configured to operate at a power rating of about 75 watts.
In some embodiments, the storage sled 1200 may also include a controller-to-controller interconnect 1242. Similar to the resource-to-resource interconnect 624 of the sled 400 discussed above, the controller-to-controller interconnect 1242 may be embodied as any type of communication interconnect capable of facilitating controller-to-controller communications. In the illustrative embodiment, the controller-to-controller interconnect 1242 is embodied as a high-speed point-to-point interconnect (e.g., faster than the I/O subsystem 622). For example, the controller-to-controller interconnect 1242 may be embodied as a QuickPath Interconnect (QPI), an UltraPath Interconnect (UPI), or other high-speed point-to-point interconnect dedicated to processor-to-processor communications.
Referring now to FIG. 13, an illustrative embodiment of the storage sled 1200 is shown. In the illustrative embodiment, the data storage 1250 is embodied as, or otherwise includes, a storage cage 1252 configured to house one or more solid state drives (SSDs) 1254. To do so, the storage cage 1252 includes a number of mounting slots 1256, each of which is configured to receive a corresponding solid state drive 1254. Each of the mounting slots 1256 includes a number of drive guides 1258 that cooperate to define an access opening 1260 of the corresponding mounting slot 1256. The storage cage 1252 is secured to the chassis-less circuit board substrate 602 such that the access openings face away from (i.e., toward the front of) the chassis-less circuit board substrate 602. As such, solid state drives 1254 are accessible while the storage sled 1200 is mounted in a corresponding rack 204. For example, a solid state drive 1254 may be swapped out of a rack 240 (e.g., via a robot) while the storage sled 1200 remains mounted in the corresponding rack 240.
The storage cage 1252 illustratively includes sixteen mounting slots 1256 and is capable of mounting and storing sixteen solid state drives 1254. Of course, the storage cage 1252 may be configured to store additional or fewer solid state drives 1254 in other embodiments. Additionally, in the illustrative embodiment, the solid state drivers are mounted vertically in the storage cage 1252, but may be mounted in the storage cage 1252 in a different orientation in other embodiments. Each solid state drive 1254 may be embodied as any type of data storage device capable of storing long term data. To do so, the solid state drives 1254 may include volatile and non-volatile memory devices discussed above.
As shown in FIG. 13, the storage controllers 1220, the communication circuit 830, and the optical data connector 834 are illustratively mounted to the top side 650 of the chassis-less circuit board substrate 602. Again, as discussed above, any suitable attachment or mounting technology may be used to mount the electrical components of the storage sled 1200 to the chassis-less circuit board substrate 602 including, for example, sockets (e.g., a processor socket), holders, brackets, soldered connections, and/or other mounting or securing techniques.
As discussed above, the individual storage controllers 1220 and the communication circuit 830 are mounted to the top side 650 of the chassis-less circuit board substrate 602 such that no two heat-producing, electrical components shadow each other. For example, the storage controllers 1220 and the communication circuit 830 are mounted in corresponding locations on the top side 650 of the chassis-less circuit board substrate 602 such that no two of those electrical components are linearly in-line with other along the direction of the airflow path 608.
The memory devices 720 of the storage sled 1200 are mounted to the bottom side 750 of the of the chassis-less circuit board substrate 602 as discussed above in regard to the sled 400. Although mounted to the bottom side 750, the memory devices 720 are communicatively coupled to the storage controllers 1220 located on the top side 650 via the I/O subsystem 622. Again, because the chassis-less circuit board substrate 602 is embodied as a double-sided circuit board, the memory devices 720 and the storage controllers 1220 may be communicatively coupled by one or more vias, connectors, or other mechanisms extending through the chassis-less circuit board substrate 602. Each of the storage controllers 1220 includes a heatsink 1270 secured thereto. As discussed above, due to the improved thermal cooling characteristics of the chassis-less circuit board substrate 602 of the storage sled 1200, none of the heatsinks 1270 include cooling fans attached thereto. That is, each of the heatsinks 1270 is embodied as a fan-less heatsink.
Referring now to FIG. 14, in some embodiments, the sled 400 may be embodied as a memory sled 1400. The storage sled 1400 is optimized, or otherwise configured, to provide other sleds 400 (e.g., compute sleds 800, accelerator sleds 1000, etc.) with access to a pool of memory (e.g., in two or more sets 1430, 1432 of memory devices 720) local to the memory sled 1200. For example, during operation, a compute sled 800 or an accelerator sled 1000 may remotely write to and/or read from one or more of the memory sets 1430, 1432 of the memory sled 1200 using a logical address space that maps to physical addresses in the memory sets 1430, 1432. The memory sled 1400 includes various components similar to components of the sled 400 and/or the compute sled 800, which have been identified in FIG. 14 using the same reference numbers. The description of such components provided above in regard to FIGS. 6, 7, and 8 apply to the corresponding components of the memory sled 1400 and is not repeated herein for clarity of the description of the memory sled 1400.
In the illustrative memory sled 1400, the physical resources 620 are embodied as memory controllers 1420. Although only two memory controllers 1420 are shown in FIG. 14, it should be appreciated that the memory sled 1400 may include additional memory controllers 1420 in other embodiments. The memory controllers 1420 may be embodied as any type of processor, controller, or control circuit capable of controlling the writing and reading of data into the memory sets 1430, 1432 based on requests received via the communication circuit 830. In the illustrative embodiment, each storage controller 1220 is connected to a corresponding memory set 1430, 1432 to write to and read from memory devices 720 within the corresponding memory set 1430, 1432 and enforce any permissions (e.g., read, write, etc.) associated with sled 400 that has sent a request to the memory sled 1400 to perform a memory access operation (e.g., read or write).
In some embodiments, the memory sled 1400 may also include a controller-to-controller interconnect 1442. Similar to the resource-to-resource interconnect 624 of the sled 400 discussed above, the controller-to-controller interconnect 1442 may be embodied as any type of communication interconnect capable of facilitating controller-to-controller communications. In the illustrative embodiment, the controller-to-controller interconnect 1442 is embodied as a high-speed point-to-point interconnect (e.g., faster than the I/O subsystem 622). For example, the controller-to-controller interconnect 1442 may be embodied as a QuickPath Interconnect (QPI), an UltraPath Interconnect (UPI), or other high-speed point-to-point interconnect dedicated to processor-to-processor communications. As such, in some embodiments, a memory controller 1420 may access, through the controller-to-controller interconnect 1442, memory that is within the memory set 1432 associated with another memory controller 1420. In some embodiments, a scalable memory controller is made of multiple smaller memory controllers, referred to herein as “chiplets”, on a memory sled (e.g., the memory sled 1400). The chiplets may be interconnected (e.g., using EMIB (Embedded Multi-Die Interconnect Bridge)). The combined chiplet memory controller may scale up to a relatively large number of memory controllers and I/O ports, (e.g., up to 16 memory channels). In some embodiments, the memory controllers 1420 may implement a memory interleave (e.g., one memory address is mapped to the memory set 1430, the next memory address is mapped to the memory set 1432, and the third address is mapped to the memory set 1430, etc.). The interleaving may be managed within the memory controllers 1420, or from CPU sockets (e.g., of the compute sled 800) across network links to the memory sets 1430, 1432, and may improve the latency associated with performing memory access operations as compared to accessing contiguous memory addresses from the same memory device.
Further, in some embodiments, the memory sled 1400 may be connected to one or more other sleds 400 (e.g., in the same rack 240 or an adjacent rack 240) through a waveguide, using the waveguide connector 1480. In the illustrative embodiment, the waveguides are 64 millimeter waveguides that provide 16 Rx (i.e., receive) lanes and 16 Rt (i.e., transmit) lanes. Each lane, in the illustrative embodiment, is either 16 Ghz or 32 Ghz. In other embodiments, the frequencies may be different. Using a waveguide may provide high throughput access to the memory pool (e.g., the memory sets 1430, 1432) to another sled (e.g., a sled 400 in the same rack 240 or an adjacent rack 240 as the memory sled 1400) without adding to the load on the optical data connector 834.
Referring now to FIG. 15, a system for executing one or more workloads (e.g., applications) may be implemented in accordance with the data center 100. In the illustrative embodiment, the system 1510 includes an orchestrator server 1520, which may be embodied as a managed node comprising a compute device (e.g., a compute sled 800) executing management software (e.g., a cloud operating environment, such as OpenStack) that is communicatively coupled to multiple sleds 400 including a large number of compute sleds 1530 (e.g., each similar to the compute sled 800), memory sleds 1540 (e.g., each similar to the memory sled 1400), accelerator sleds 1550 (e.g., each similar to the memory sled 1000), and storage sleds 1560 (e.g., each similar to the storage sled 1200). One or more of the sleds 1530, 1540, 1550, 1560 may be grouped into a managed node 1570, such as by the orchestrator server 1520, to collectively perform a workload (e.g., an application 1532 executed in a virtual machine or in a container). The managed node 1570 may be embodied as an assembly of physical resources 620, such as processors 820, memory resources 720, accelerator circuits 1020, or data storage 1250, from the same or different sleds 400. Further, the managed node may be established, defined, or “spun up” by the orchestrator server 1520 at the time a workload is to be assigned to the managed node or at any other time, and may exist regardless of whether any workloads are presently assigned to the managed node. In the illustrative embodiment, the orchestrator server 1520 may selectively allocate and/or deallocate physical resources 620 from the sleds 400 and/or add or remove one or more sleds 400 from the managed node 1570 as a function of quality of service (QoS) targets (e.g., performance targets associated with a throughput, latency, instructions per second, etc.) associated with a service level agreement for the workload (e.g., the application 1532). In doing so, the orchestrator server 1520 may receive telemetry data indicative of performance conditions (e.g., throughput, latency, instructions per second, etc.) in each sled 400 of the managed node 1570 and compare the telemetry data to the quality of service targets to determine whether the quality of service targets are being satisfied. If the so, the orchestrator server 1520 may additionally determine whether one or more physical resources may be deallocated from the managed node 1570 while still satisfying the QoS targets, thereby freeing up those physical resources for use in another managed node (e.g., to execute a different workload). Alternatively, if the QoS targets are not presently satisfied, the orchestrator server 1520 may determine to dynamically allocate additional physical resources to assist in the execution of the workload (e.g., the application 1532) while the workload is executing
Additionally, in some embodiments, the orchestrator server 1520 may identify trends in the resource utilization of the workload (e.g., the application 1532), such as by identifying phases of execution (e.g., time periods in which different operations, each having different resource utilizations characteristics, are performed) of the workload (e.g., the application 1532) and pre-emptively identifying available resources in the data center 100 and allocating them to the managed node 1570 (e.g., within a predefined time period of the associated phase beginning). In some embodiments, the orchestrator server 1520 may model performance based on various latencies and a distribution scheme to place workloads among compute sleds and other resources (e.g., accelerator sleds, memory sleds, storage sleds) in the data center 100. For example, the orchestrator server 1520 may utilize a model that accounts for the performance of resources on the sleds 400 (e.g., FPGA performance, memory access latency, etc.) and the performance (e.g., congestion, latency, bandwidth) of the path through the network to the resource (e.g., FPGA). As such, the orchestrator server 1520 may determine which resource(s) should be used with which workloads based on the total latency associated with each potential resource available in the data center 100 (e.g., the latency associated with the performance of the resource itself in addition to the latency associated with the path through the network between the compute sled executing the workload and the sled 400 on which the resource is located).
In some embodiments, the orchestrator server 1520 may generate a map of heat generation in the data center 100 using telemetry data (e.g., temperatures, fan speeds, etc.) reported from the sleds 400 and allocate resources to managed nodes as a function of the map of heat generation and predicted heat generation associated with different workloads, to maintain a target temperature and heat distribution in the data center 100. Additionally or alternatively, in some embodiments, the orchestrator server 1520 may organize received telemetry data into a hierarchical model that is indicative of a relationship between the managed nodes (e.g., a spatial relationship such as the physical locations of the resources of the managed nodes within the data center 100 and/or a functional relationship, such as groupings of the managed nodes by the customers the managed nodes provide services for, the types of functions typically performed by the managed nodes, managed nodes that typically share or exchange workloads among each other, etc.). Based on differences in the physical locations and resources in the managed nodes, a given workload may exhibit different resource utilizations (e.g., cause a different internal temperature, use a different percentage of processor or memory capacity) across the resources of different managed nodes. The orchestrator server 1520 may determine the differences based on the telemetry data stored in the hierarchical model and factor the differences into a prediction of future resource utilization of a workload if the workload is reassigned from one managed node to another managed node, to accurately balance resource utilization in the data center 100.
To reduce the computational load on the orchestrator server 1520 and the data transfer load on the network, in some embodiments, the orchestrator server 1520 may send self-test information to the sleds 400 to enable each sled 400 to locally (e.g., on the sled 400) determine whether telemetry data generated by the sled 400 satisfies one or more conditions (e.g., an available capacity that satisfies a predefined threshold, a temperature that satisfies a predefined threshold, etc.). Each sled 400 may then report back a simplified result (e.g., yes or no) to the orchestrator server 1520, which the orchestrator server 1520 may utilize in determining the allocation of resources to managed nodes.
Referring now to FIG. 16, a system 1610, similar to the system 1510, for managing a kernel topology for a workload may be implemented in accordance with the data center 100. In an example embodiment, the system 1610 includes an orchestrator server 1620 communicatively coupled with multiple sleds, including a compute sled 1630 and accelerator sleds 1640, 1650 and 1660. The compute sled 1630 and accelerator sleds 1640, 1650 and 1660 may be grouped into a managed node, such as by the orchestrator server 1620. The managed node may collectively execute a workload, such as an application (e.g., application 1634). A managed node may be embodied as an assembly of resources (e.g., physical resources), such as compute resources, memory resources, storage resources, or other resources (e.g., accelerator resources), from the same or different sleds or racks. Further, a managed node may be established, defined, or “spun up” by the orchestrator server 1620 at the time a workload is to be assigned to the managed node or at any other time, and may exist regardless of whether any workloads are presently assigned to the managed node. The system 1610 may be located in a data center and provide storage and compute services (e.g., cloud services) to a client device 1614 that is in communication with the system 1610 through a network 1612. The orchestrator server 1620 may support a cloud operating environment, such as OpenStack, and managed nodes established by the orchestrator server 1620 may execute one or more applications or processes (i.e., workloads), such as in virtual machines or containers, on behalf of a user of the client device 1614.
Illustratively, the compute sled 1630 includes one or more central processing units (CPUs) 1632 (e.g., a processor or other device or circuitry capable of performing a series of operations) that executes a workload (e.g., application 1634). The accelerator sled 1640 includes an accelerator device 1642. Additionally, the compute sled 1650 includes accelerator devices 1652 and 1656. The accelerator sled 1660 includes accelerator devices 1662 and 1667. Each of the accelerator devices 1642, 1652, 1656, 1662, and 1667 may be embodied as any device or circuitry usable to accelerate the execution of one or more operations. For example, the accelerator devices described herein may be embodied as any device or circuitry (e.g., a specialized processor, a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a graphics processing unit (GPU), reconfigurable hardware, etc.) capable of accelerating execution of a portion of the workload, such as a workload task (e.g., a set of operations within a workload). Further, each of the accelerator devices are configured with accelerated kernels. Illustratively, the accelerator device 1642 includes kernel A 1644, kernel B 1645; the accelerator device 1652 includes kernel A 1653 and kernel B 1654; the accelerator device 1656 includes kernel A 1657 and kernel B 1658; the accelerator device 1662 includes kernel A 1663, kernel B 1664, kernel C 1665, and kernel D 1666; and the accelerator device 1667 includes kernel A 1668, kernel B 1669, kernel C 1670, and kernel D 1671. Each of the accelerated kernels may be embodied as a set of code or a configuration of a portion of the corresponding accelerator device that causes the respective accelerator device to perform one or more accelerated functions (e.g., cryptographic operations, compression operations, etc.).
Each of the accelerator sleds 1640, 1650, and 1660 provide accelerated functions as a service for workloads processed by the managed node. In particular, each accelerator sled 1640, 1650, and 1660 may process requests from other sleds within the managed node (e.g., the compute sled 1630) to accelerate a function. For instance, FIG. 16 depicts the compute sled 1630 executing the application 1634. The application 1634 may include functions to be performed in sequence. The compute sled 1630 may send a request to the accelerator sleds to accelerate the execution of each function, thereby offloading the execution of the function to an accelerator device residing on the accelerator sled. The accelerator sled may, in response to the request, provision a kernel on the accelerator device. For example, the accelerator sled may load a bit stream (e.g., any data indicative of code executable by an accelerator device to configure itself with a corresponding kernel) associated with the kernel into a slot (e.g., a subset of circuitry or other logic units) of the accelerator device. The application 1634 may include a variety of functions, such as cryptographic operations, machine learning algorithms, and the like. The kernel provisioned on the accelerator device may be suited to accelerate the execution of the function. For example, the underlying function may involve matrix multiplication. The kernel provisioned with the accelerator device may be specific to processing matrix multiplication operations. Once the kernel completes acceleration of the function, the kernel may return resulting data to the compute sled 1630. The orchestrator server 1620 may track (e.g., via a database) which kernels are registered to which accelerator sleds and which accelerator devices.
As further described herein, the orchestrator server 1620 may define a topology for kernel-to-kernel communication as a function of a workload (e.g., the application 1634) across the system 1610. More particularly, the topology defines communication links between a given kernel and one or more kernels based on a flow of data in a workload, such as from task to task in a workload. For example, a kernel A (e.g., kernel A 1644) may be mapped, via a topology, to a kernel C (e.g., kernel C 1665). Kernel A 1644 may accelerate a function associated with a task and output the resulting data to kernel C 1665, which in turn accelerates a function associated with another task, using the data output from kernel A. A given kernel may be mapped to another kernel within the same accelerator device, within the same sled, between slots (e.g., subsets of circuitry or other logic units) in the same accelerator device, etc. Further, the mappings may be based on various characteristics, such as processing requirements, size of the data being processed in a task, a location of the accelerator device or sled relative to the compute sled executing a workload, and the like.
Further still, the orchestrator server 1620 may scale accelerated kernel resources according to the kernel topology. In particular, the orchestrator server 1620 may, as a function of the resource utilization, re-provision a kernel on a number of accelerator devices in the system 1610. Continuing the previous example, the kernel C may be provisioned on accelerator devices 1662 and 1667 for a workload task. Further, incoming traffic to the accelerator devices 1662 and 1667 may exceed a specified threshold (e.g., based on a policy). The orchestrator server 1620, in response, may provision additional instances of kernel C on other accelerator devices, such as accelerator devices 1642 or 1652. Conversely, if incoming traffic to accelerator devices 1662 and 1667 falls below a specified threshold, the orchestrator server 1620, in response, may de-provision the kernel C in one of the accelerator devices 1662 or 1667. Advantageously, scaling a kernel configuration enables more efficient usage of the accelerator devices in the system 1610 as compared to typical systems.
In addition, the orchestrator server 1620 may scale accelerated kernel resources in response to detecting that a given accelerator device is in a critical power state. For instance, each accelerator device may report power consumption to the orchestrator server 1620. The orchestrator server 1620 may determine that power consumption in an accelerator device configured with a given kernel exceeds a specified threshold. In response, the orchestrator server 1620 may identify an accelerator device in the system that is available (e.g., is at a suitable power consumption level) to be provisioned with the kernel. Once identified, the orchestrator server 1620 may provision the kernel on the accelerator device and update the kernel topology. As a result, power consumption by the kernel is partially shifted to the additional accelerator device, reducing power consumption in the accelerator device previously in a critical power state.
In addition, while the orchestrator server 1620 is configured to provision kernels on accelerator devices in the system 1610, an accelerator sled (e.g., accelerator sleds 1640, 1650, or 1660) that is configured with a kernel as part of the topology may store a kernel bit stream associated with another kernel to which the configured kernel is mapped. The underlying workload (e.g., application 1634) may identify that further additional accelerator devices should be configured with the kernel (e.g., as a function of observed telemetry data). As further described herein, the accelerator sled may include logic to provision an accelerator device (e.g., on another accelerator sled) with the other kernel. To do so, the accelerator sled may broadcast a request packet to accelerator devices to determine which of the devices are available (e.g., currently not configured with a kernel, has resources to accommodate execution by the kernel, etc.). In response, one or more of the accelerator devices may return a response to the accelerator sled. The accelerator sled may identify which of the accelerator devices to provision as a function of one or more characteristics of each of the available devices, such as current resources available on the device. Once identified, the kernel may encapsulate the stored bit stream of the other kernel in one or more provisioning packets and send the provisioning packets to the accelerator device.
Referring now to FIG. 17, the orchestrator server 1620 may be embodied as any type of compute device capable of performing the functions described herein, including receiving a kernel configuration request to provision kernels on multiple accelerator devices, determining, as a function of one or more requirements of a workload, a topology of kernels to service the request, and configure the kernels with the accelerator devices according to the determined topology. The orchestrator server 1620 may also be embodied as any type of compute device capable of detecting whether an accelerator device configured with a kernel is in a critical power state, determining another accelerator device that is available to be configured with the kernel, and configuring, in response to a determination that the accelerator device is in a critical power state, the other accelerator device with the kernel, where the kernel configured with the accelerator device is to share acceleration of execution of a workload with the kernel configured with the other accelerator device.
As shown in FIG. 17, the illustrative orchestrator server 1620 includes a compute engine 1702, an input/output (I/O) subsystem 1708, communication circuitry 1710, and one or more data storage devices 1714. Of course, in other embodiments, the orchestrator server 1620 may include other or additional components, such as those commonly found in a computer (e.g., display, peripheral devices, etc.). Additionally, in some embodiments, one or more of the illustrative components may be incorporated in, or otherwise form a portion of, another component.
The compute engine 1702 may be embodied as any type of device or collection of devices capable of performing various compute functions described below. In some embodiments, the compute engine 1702 may be embodied as a single device such as an integrated circuit, an embedded system, a FPGA, a system-on-a-chip (SoC), or other integrated system or device. Additionally, in some embodiments, the compute engine 1702 includes or is embodied as a processor 1704 and a memory 1706. The processor 1704 may be embodied as any type of processor capable of performing the functions described herein. For example, the processor 1704 may be embodied as a single or multi-core processor(s), a microcontroller, or other processor or processing/controlling circuit. In some embodiments, the processor 1704 may be embodied as, include, or be coupled to an FPGA, an ASIC, reconfigurable hardware or hardware circuitry, or other specialized hardware to facilitate performance of the functions described herein.
The memory 1706 may be embodied as any type of volatile (e.g., dynamic random access memory (DRAM), etc.) or non-volatile memory or data storage capable of performing the functions described herein. Volatile memory may be a storage medium that requires power to maintain the state of data stored by the medium. Non-limiting examples of volatile memory may include various types of random access memory (RAM), such as DRAM or static random access memory (SRAM). One particular type of DRAM that may be used in a memory module is synchronous dynamic random access memory (SDRAM). In particular embodiments, DRAM of a memory component may comply with a standard promulgated by JEDEC, such as JESD79F for DDR SDRAM, JESD79-2F for DDR2 SDRAM, JESD79-3F for DDR3 SDRAM, JESD79-4A for DDR4 SDRAM, JESD209 for Low Power DDR (LPDDR), JESD209-2 for LPDDR2, JESD209-3 for LPDDR3, and JESD209-4 for LPDDR4 (these standards are available at www.jedec.org). Such standards (and similar standards) may be referred to as DDR-based standards and communication interfaces of the storage devices that implement such standards may be referred to as DDR-based interfaces.
In one embodiment, the memory device is a block addressable memory device, such as those based on NAND or NOR technologies. A memory device may also include future generation nonvolatile devices, such as a three dimensional crosspoint memory device (e.g., Intel 3D XPoint™ memory), or other byte addressable write-in-place nonvolatile memory devices. In one embodiment, the memory device may be or may include memory devices that use chalcogenide glass, multi-threshold level NAND flash memory, NOR flash memory, single or multi-level Phase Change Memory (PCM), a resistive memory, nanowire memory, ferroelectric transistor random access memory (FeTRAM), anti-ferroelectric memory, magnetoresistive random access memory (MRAM) memory that incorporates memristor technology, resistive memory including the metal oxide base, the oxygen vacancy base and the conductive bridge Random Access Memory (CB-RAM), or spin transfer torque (STT)-MRAM, a spintronic magnetic junction memory based device, a magnetic tunneling junction (MTJ) based device, a DW (Domain Wall) and SOT (Spin Orbit Transfer) based device, a thyristor based memory device, or a combination of any of the above, or other memory. The memory device may refer to the die itself and/or to a packaged memory product.
In some embodiments, 3D crosspoint memory (e.g., Intel 3D XPoint™ memory) may comprise a transistor-less stackable cross point architecture in which memory cells sit at the intersection of word lines and bit lines and are individually addressable and in which bit storage is based on a change in bulk resistance. In some embodiments, all or a portion of the memory 1706 may be integrated into the processor 1704. In operation, the memory 1706 may store various software and data used during operation.
The compute engine 1702 is communicatively coupled with other components of the orchestrator server 1620 via the I/O subsystem 1708, which may be embodied as circuitry and/or components to facilitate input/output operations with the compute engine 1702 (e.g., with the processor 1704 and/or the memory 1706) and other components of the orchestrator server 1620. For example, the I/O subsystem 1708 may be embodied as, or otherwise include, memory controller hubs, input/output control hubs, integrated sensor hubs, firmware devices, communication links (e.g., point-to-point links, bus links, wires, cables, light guides, printed circuit board traces, etc.), and/or other components and subsystems to facilitate the input/output operations. In some embodiments, the I/O subsystem 1708 may form a portion of a system-on-a-chip (SoC) and be incorporated, along with one or more of the processor 1704, the memory 1706, and other components of the orchestrator server 1620, into the compute engine 1702.
The communication circuitry 1710 may be embodied as any communication circuit, device, or collection thereof, capable of enabling communications over the network 1612 between the orchestrator server 1620 and another compute device (e.g., the compute sled 1630, the accelerator sleds 1640, 1650, and 1660, etc.). The communication circuitry 1710 may be configured to use any one or more communication technology (e.g., wired or wireless communications) and associated protocols (e.g., Ethernet, Bluetooth®, Wi-Fi®, WiMAX, etc.) to effect such communication.
The illustrative communication circuitry 1710 includes a network interface controller (NIC) 1712, which may also be referred to as a host fabric interface (HFI). The NIC 1712 may be embodied as one or more add-in-boards, daughter cards, network interface cards, controller chips, chipsets, or other devices that may be used by the orchestrator server 1620 to connect with another compute device (e.g., the compute sled 1630, the accelerator sleds 1640, 1650, and 1660, etc.). In some embodiments, the NIC 1712 may be embodied as part of a system-on-a-chip (SoC) that includes one or more processors, or included on a multichip package that also contains one or more processors. In some embodiments, the NIC 1712 may include a local processor (not shown) and/or a local memory (not shown) that are both local to the NIC 1712. In such embodiments, the local processor of the NIC 1712 may be capable of performing one or more of the functions of the compute engine 1702 described herein. Additionally or alternatively, in such embodiments, the local memory of the NIC 1712 may be integrated into one or more components of the orchestrator server 1620 at the board level, socket level, chip level, and/or other levels.
The one or more illustrative data storage devices 1714 may be embodied as any type of devices configured for short-term or long-term storage of data such as, for example, memory devices and circuits, memory cards, hard disk drives (HDDs), solid-state drives (SSDs), or other data storage devices. Each data storage device 1714 may include a system partition that stores data and firmware code for the data storage device 1714. Each data storage device 1714 may also include an operating system partition that stores data files and executables for an operating system. Additionally or alternatively, the orchestrator server 1620 may include one or more peripheral devices 1716. Such peripheral devices 1716 may include any type of peripheral device commonly found in a compute device such as a display, speakers, a mouse, a keyboard, and/or other input/output devices, interface devices, and/or other peripheral devices.
Referring now to FIG. 18, an accelerator sled 1800 may be embodied as any type of compute device capable of performing the functions described herein, including receiving a request by a workload to provision a kernel on an accelerator device, identifying an accelerator device on which to provision the kernel, and configuring the identified accelerator device with the kernel. The accelerator sled 1800 may be representative of any of the accelerator sleds 1640, 1650, or 1660 depicted in FIG. 16. As shown in FIG. 18, the accelerator sled 1800 includes a compute engine 1802, an I/O subsystem 1808, communication circuitry 1810, one or more data storage devices 1814, and one or more accelerator devices 1818. Of course, in other embodiments, the accelerator sled 1800 may include other or additional components, such as those commonly found in a computer (e.g., display, peripheral devices, etc.). Additionally, in some embodiments, one or more of the illustrative components may be incorporated in, or otherwise form a portion of, another component.
The compute engine 1802 may be embodied as any type of device or collection of devices capable of performing various compute functions described below and is similar to the compute engine 1802 of FIG. 17. The memory 1806 may be embodied as any type of volatile (e.g., DRAM, etc.) or non-volatile memory or data storage capable of performing the functions described herein. In operation, the memory 1806 may store various software and data used during operation such as kernel topology data, telemetry data, kernel bit stream data, applications, programs, libraries, and drivers. The I/O subsystem 1808 is similar to the I/O subsystem 1708 described with reference to FIG. 17. The communication circuitry 1810, which, in the illustrative embodiment, includes a NIC 1812, is similar to the communication circuitry 1710 and NIC 1712 described with reference to FIG. 17. Additionally, the data storage devices 1814 are similar to the data storage devices 1714 described with reference to FIG. 17. Further, the peripheral devices 1816 are similar to the peripheral devices 1716, described with reference to FIG. 17.
The accelerator devices 1818 can be representative of accelerator devices in the system 1610 depicted in FIG. 16, such as any combination of accelerator devices 1642, 1652, 1656, 1662, and 1667. The accelerator devices 1818 may form an accelerator subsystem that includes one or more buses or other interfaces between the accelerator devices in the accelerator sled 1800 to enable the accelerator devices to share data. Further, each accelerator device 1810 may send data via the NIC 1812 to other accelerator devices in the system 1610, based on a kernel topology defined by the orchestrator server 1620. Each accelerator device 1818 may be embodied as any device or circuitry (e.g., a specialized processor, an FPGA, an ASIC, a GPU, reconfigurable hardware, etc.) capable of accelerating the execution of a function.
The client device 1614 and the compute sled 1630 may have components similar to those described in FIG. 18. The description of those components of the orchestrator server 1620 and the accelerator sled 1800 is equally applicable to the description of components of those devices and is not repeated herein for clarity of the description. Further, it should be appreciated that any of the client device 1614, the orchestrator server 1620, and the sleds 1630, 1640, 1650, an 1660 may include other components, sub-components, and devices commonly found in a computing device, which are not discussed above in reference to the orchestrator server 1620 and the accelerator sled 1800 and not discussed herein for clarity of the description.
As described above, the client device 1614, the orchestrator server 1620, and the sleds 1630, 1640, 1650, and 1660 are illustratively in communication via the network 1612, which may be embodied as any type of wired or wireless communication network, including global networks (e.g., the Internet), local area networks (LANs) or wide area networks (WANs), cellular networks (e.g., Global System for Mobile Communications (GSM), 3G, Long Term Evolution (LTE), Worldwide Interoperability for Microwave Access (WiMAX), etc.), digital subscriber line (DSL) networks, cable networks (e.g., coaxial networks, fiber networks, etc.), or any combination thereof.
Referring now to FIG. 19, the orchestrator server 1620 may establish an environment 1900 during operation. Illustratively, the environment 1900 includes a network communicator 1920 and a topology manager 1930. Each of the components of the environment 1900 may be embodied as hardware, firmware, software, or a combination thereof. As such, in some embodiments, one or more of the components of the environment 1900 may be embodied as circuitry or a collection of electrical devices (e.g., network communicator circuitry 1920, topology manager circuitry 1930, etc.). It should be appreciated that, in such embodiments, one or more of the network communicator circuitry 1920 or the topology manager circuitry 1930 may form a portion of one or more of the compute engine 1702, the communication circuitry 1710, the I/O subsystem 1708, and/or other components of the orchestrator server 1620.
In the illustrative embodiment, the environment 1900 includes kernel topology data 1902, which may be embodied as any data indicative of kernel-to-kernel mappings for a workload. The kernel topology 1902 may indicate a flow of data originating from the underlying application to one or more kernels in sequence and flowing back to the application. For example, a kernel A may accelerate a function using data provided by the application and send the result of the acceleration to a kernel B. In turn, the kernel B accelerates a function using the resulting data sent by kernel A as input, and so on. The kernel topology data 1902 is also indicative of kernel configurations on accelerator devices in the system 1610. For example, kernel topology data 1902 may include data specifying that kernel A is configured on accelerator devices 1642, 1652, 1656, 1662, and 1667. Such data may be indicative of a network address of the accelerator device, a network address of the accelerator sled, a slot location on the accelerative device, and the like. A mapping between a given kernel and another kernel may also indicate any intermediary NICs or switch devices connecting the kernels with one another, such as providing address data of each NIC or switch device. In some embodiments, the kernels may be interconnected via an accelerator subsystem interface that connects components of an accelerator device, including kernels, with one another to form a kernel-to-kernel network. The accelerator subsystem interface may expose a virtual address space that allows kernels to identify and communicate with one another in the network.
The environment 1900 also includes policy data 1904, which may be embodied as any data indicative of conditions, which, when triggered, cause the orchestrator server 1620 to scale kernel instances in one or more of the accelerator devices of the system 1610. The orchestrator server 1620 may, in response to a condition trigger, provision additional instances of a kernel inside the same accelerator device on which the kernel is presently configured, such as in another slot of the accelerator device. In addition, the orchestrator server 1620 may provision additional instances of a kernel inside the same accelerator sled on which the kernel is presently configured, such as in another accelerator device residing on the accelerator sled. In addition, the orchestrator server 1620 may provision additional instances of a kernel on an accelerator device of another accelerator sled in the system 1610. Further, the orchestrator server 1620 may remove kernel instances from an accelerator device after a condition is triggered (e.g., scenarios in which resource utilization falls below a specified threshold). Policies may specify that the accelerator device on which the kernel is provisioned be determined as a function of characteristics of the accelerator device, such as the availability of the accelerator device (e.g., whether the accelerator device has resources to be configured with the kernel and utilization to process workload data), the location of the accelerator device relative to the compute sled executing the workload, the location of the accelerator device relative to other kernels linked to the kernel according to the kernel topology data 1902, and the like.
The environment 1900 also includes telemetry data 1906, which may be embodied as any data indicative of observed performance of accelerator sleds and accelerator devices in the system 1610. Telemetry may include metrics such as power consumption of a given sled or device, amount of kernel connections, latency, average period of connections, amount of data transferred per connection, etc. The orchestrator server 1620 may receive telemetry data 1906 from each accelerator sled. Alternatively or in addition, an external resource monitor may observe usage in the accelerator sleds and report the corresponding telemetry data 1906 to the orchestrator server 1620.
Illustratively, the network communicator 1920, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to facilitate inbound and outbound network communications (e.g., network traffic, network packets, network flows, etc.) to and from the orchestrator server 1620, respectively. To do so, the network communicator 1920 is configured to receive and process data packets from one system or computing device (e.g., the compute sled 1630) and to prepare and send data packets to another computing device or system (e.g., the accelerator sleds 1640, 1650, and 1660). Accordingly, in some embodiments, at least a portion of the functionality of the network communicator 1920 may be performed by the communication circuitry 1710, and, in the illustrative embodiment, by the NIC 1712.
The topology manager 1930, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof, is configured to identify available accelerator device resources, determine kernel topology data 1902 as a function of a workload and the available accelerator device resources, and provision accelerated kernels on the accelerator devices according to the kernel topology data 1902. The topology manager 1930 also scales kernel instances on the accelerator devices as a function of observed telemetry data 1906. As shown, the topology manager 1930 includes a resource discovery component 1932, a provision component 1934, and a monitor component 1936.
In one embodiment, the resource discovery component 1932 is configured to identify resource availability in accelerator devices of the system 1610, such as in accelerator devices 1642, 1652, 1656, 1662, and 1667. For example, the resource discovery component 1932 may query each accelerator device for a present resource utilization or other characteristics, such a total amount of slots on the device, an amount of slots configured with kernels on the device, a device location, and the like. The resource discovery component 1932 may do similar for the accelerator sleds on which the accelerator devices reside. For example, the resource discovery component 1932 may query an accelerator sled for available accelerator devices therein. The resource discovery component 1932 also determines kernel bit streams applicable to a given workload. For example, a given workload may be configured to execute using proprietary kernels designed for the workload or other third-party kernels that are targeted to performing functions associated with one or more workload tasks. A compute sled executing a workload may send data describing the workload to the orchestrator server 1620, and in turn, the resource discovery component 1932 may determine suitable kernels for the workload. Further, the resource discovery component 1932 may determine a storage location of each kernel bit stream used for the workload.
In the illustrative embodiment, the provision component 1934 is configured to generate a kernel topology as a function of the available resources in the system 1610 and workload characteristics (e.g., resource requirements of the workload, functions included with each workload task, and the like). For example, to do so, the provision component 1934 may evaluate the location of compute sled executing a given workload and determine available accelerator devices that are located relatively near (e.g., within a predefined distance) to the compute sled for satisfying a latency target. The provision component 1934 may also evaluate an amount of processing power required for a given workload task and identify which of the available resources is capable of providing such processing power. The provision component 1934 can determine, based on such evaluation, a configuration and routing of kernels on the accelerator sleds. The provision component 1934 is further configured to load kernel bit stream data in one or more accelerator devices according to the kernel topology. For example, the provision component 1934 may obtain the kernel bit stream data from a previously identified storage location and send the bit stream data to a given accelerator device. More specifically, the provision component 1934 may send the bit stream to the accelerator sled on which the accelerator device resides, and in turn, the accelerator sled may program the bit stream in a designated slot of the accelerator device. The provision component 1934 may write and maintain a record indicative of the location of the kernel in the kernel topology data 1902. The record may also indicate one or more kernels to which to send data and one or more kernels from which to receive data. The record may also indicate other accelerator devices that are configured with the kernel.
In one embodiment, the monitor component 1936 is configured to collect telemetry data from accelerator sleds (e.g., accelerator sleds 1640, 1650, and 1660) and the devices residing thereon. The monitor component 1936 may do so in a variety of manners. For example, the monitor component 1936 may regularly query a given accelerator sled for telemetry data. As another example, the accelerator subsystem interface described above may provide metrics in each accelerator device at specified intervals. As another example, an external monitor may collect telemetry data from each accelerator sled and send the collected telemetry data to the monitor component 1936. The monitor component 1936 may store the telemetry data on the orchestrator server 1620 in a data store local to the server (e.g., as telemetry data 1906).
In one embodiment, the provision component 1934 is further configured to scale kernel instances in accelerator devices for a given workload as a function of the telemetry data 1906 and the policy data 1904. For example, the telemetry data 1906 may indicate that an amount of connections between one kernel and another kernel is relatively high. Further, the policy data 1904 may indicate that if an amount of connections exceeds a specified threshold, the amount of instances for one or both of the kernels should be increased. In response, the provision component 1934 may identify one or more accelerator devices that satisfy requirements of policy data 1904 for provisioning additional kernel instances. Once identified, the provision component 1934 may configure the kernel instances on the accelerator device (e.g., by sending kernel bit stream data to the accelerator device and causing the accelerator device to load the kernel bit stream). The provision component 1934 may be configured to update the kernel topology data 1902 to include the added kernel instance to the accelerator device. Further, in the illustrative embodiment, the provision component 1934 may scale kernel instances on additional accelerator devices to shift power consumption across the additional accelerator devices. For instance, the policy data 1904 may include a scaling condition that is triggered when power consumption in a given accelerator device (or sled) exceeds a specified threshold or otherwise reaches a critical power state.
It should be appreciated that each of the resource discovery component 1932, provision component 1934, and monitor component 1936 may be separately embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof. For example, the discovery component 1932 may be embodied as a hardware component, while the provision component and the monitor component 1936 are embodied as virtualized hardware components or as some other combination of hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof.
Referring now to FIG. 20, the accelerator sled 1800 may establish an environment 2000 during operation. Of course, any of the accelerator sleds 1640, 1650, 1660 may similarly establish the environment 2000 during operation. Illustratively, the environment 2000 includes a network communicator 2020 and an accelerator device manager 2030. Each of the components of the environment 2000 may be embodied as hardware, firmware, software, or a combination thereof. As such, in some embodiments, one or more of the components of the environment 2000 may be embodied as circuitry or a collection of electrical devices (e.g., network communicator circuitry 2020, accelerator device manager circuitry 2030, etc.). It should be appreciated that, in such embodiments, one or more of the network communicator circuitry 2020 or accelerator device manager circuitry 2030 may form a portion of one or more of the compute engine 1802, the communication circuitry 1810, the I/O subsystem 1808, and/or other components of the accelerator sled 1800. As shown, the environment 2000 includes kernel topology data 2002, which may be embodied as any data indicative of mappings of kernel configurations in the system 1610, similar to that of the kernel topology data 1902 described above. Further, the environment 2000 includes telemetry data 2004, which may be embodied as any data indicative of observed performance of the accelerator sled 1800 and accelerator devices 1818 (e.g., power consumption, amount of kernel connections, latency, average period of connections, amount of data transferred per connection, etc.). Further still, the environment 2000 includes kernel bit stream data 2006, which may be embodied as any data indicative of a predefined bit stream representative of a kernel for accelerating one or more functions of a workload (e.g., machine learning algorithms, cryptographic operations, compression functions, and so on). The kernel bit stream data 2006 includes bit streams of kernels not configured within an accelerator device 1818 but may be representative of a kernel that sends or receives data to a kernel configured with the accelerator device 1818 (also referred to herein as a “target kernel”).
The network communicator 2020, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to facilitate inbound and outbound network communications (e.g., network traffic, network packets, network flows, etc.) to and from the accelerator sled 1800, respectively. To do so, the network communicator 2020 is configured to receive and process data packets from one system or computing device (e.g., the orchestrator server 1620) and to prepare and send data packets to another computing device or system (e.g., the compute sled 1630, or other accelerator devices 1640, 1650, or 1660). Accordingly, in some embodiments, at least a portion of the functionality of the network communicator 2020 may be performed by the communication circuitry 1810, and, in the illustrative embodiment, by the NIC 1812.
The accelerator device manager 2030, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof, is configured to receive a request by a workload to provision a kernel on one of the accelerator devices 1818, identify one of the accelerator devices 1818 on which to provision the kernel, and configure the identified accelerator device with the kernel. As shown, the accelerator device manager 2030 includes a utilization manager 2034 and a kernel manager 2036.
In one embodiment, the utilization manager 2034 is configured to evaluate kernel-to-kernel activity between a kernel configured with the accelerator devices 1818 and target kernels connected with the kernel via the kernel topology data 2002. For instance, the utilization manager 2034 may observe telemetry data 2004 relating to the accelerator devices 1818 with which the kernel is configured and determine, as a function of observed telemetry between the kernels (e.g., communication latency, an amount of connections established between the kernels within a given time period, etc.), whether to provision additional instances of target kernels on another accelerator device in the system 1610. For example, the utilization manager 2034 may determine to provision additional instances in the event that latency between kernels exceeds a specified threshold.
In one embodiment, the kernel manager 2036 is configured to provision kernel bit stream data 2006 indicative of a target kernel on one or more accelerator devices in the system 1610. To do so, the kernel manager 2036 may broadcast a request packet to accelerator devices in the system 1610. The packet may include data indicative of a request for a response from accelerator devices that are not presently configured with any kernel (or are otherwise not presently in use). The kernel manager 2036 may then receive a response to the packet from such accelerator devices. The response may include characteristics of the accelerator device, such as a device location, total number of slots on the accelerator device, total number of available slots (e.g., those not presently being used to perform an accelerated function), and the like. The kernel manager 2036 may, based on the characteristics, determine which of the accelerator devices to provision with the target kernel.
Once determined, the kernel manager 2036 may retrieve a bit stream associated with the target kernel and encapsulate the bit stream in one or more provisioning packets. The kernel manager 2036 may then send the provisioning packets to the accelerator sled in which the accelerator device resides. The provisioning packets may include additional data that directs a destination accelerator sled to load the bit stream into one of the slots of the accelerator device. In response, the receiving accelerator sled loads the kernel. The kernel manager 2036 may update the kernel topology data 2002 to indicate that the target kernel is loaded on the additional accelerator device. Further, the kernel manager 2036 notifies the orchestrator server 1620 that the additional accelerator device has been provisioned. In turn, the orchestrator server 1620 may update the kernel topology data 1902 as a function of the notification.
Referring now to FIG. 21, the orchestrator server 1620, in operation, may execute a method 2100 to determine a topology of accelerated kernels in the system 1610. As shown, the method 2100 begins with block 2102, where the orchestrator server 1620 receives a kernel configuration request from a compute sled (e.g., the compute sled 1630) executing a workload. For example, the orchestrator server 1620 may receive the request at initialization of the workload. The request may include several resource requirements of the workload, a description of each task in the workload, and functions included in each task.
In block 2104, the orchestrator server 1620 determines one or more kernel topologies as a function of the workload requirements provided with the request. In particular, in block 2106, the orchestrator server 1620 determines characteristics of each accelerator device in the system 1610. Example characteristics include a power state, total amount of slots available, present amount of kernels loaded in each slot, a location of the device, and the like. In block 2108, the orchestrator server 1620 determines one or more configurations of accelerator devices with the kernels as a function of characteristics that satisfy the workload requirements. For example, the workload requirements may require that any function to be accelerated should be performed by an accelerator device in a sled that is located relatively near (e.g., within a predefined distance of) the compute sled executing the workload. The orchestrator server 1620 may determine, based on such a requirement, to load kernels on accelerator devices that satisfy those requirements.
In block 2110, the orchestrator server 1620 may optionally return the determined kernel topologies to the requesting compute sled for selection. In turn, the compute sled may determine a kernel topology to select based on the workload requirements and any additional criteria. In addition, a user may also make a selection of one of the determined kernel topologies. In block 2112, the orchestrator server 1620 receives the selection. In block 2114, the orchestrator server 1620 configures the accelerator devices according to the determined kernel topology. In particular, in block 2116, the orchestrator server 1620 identifies, based on characteristics of each accelerator device as provided in the kernel topology, one or more accelerator devices to provision. The orchestrator server 1620 may identify the network address of each accelerator device to provision.
In block 2118, the orchestrator server 1620 performs the following for each identified accelerator device. In block 2120, the orchestrator server 1620 retrieves the kernel bit streams to provision on the accelerator device according to the kernel topology. For example, the orchestrator server 1620 may retrieve the bit streams from a data store of kernel bit streams associated with the workload. In block 2122, the orchestrator server 1620 loads the bit streams on the accelerator device. More particularly, the orchestrator server 1620 may transmit the kernel bit stream to the accelerator sled. The accelerator sled may include logic to configure the kernel bit stream on a target accelerator device when received from the orchestrator server 1620. In other embodiments, the orchestrator server 1620 may encapsulate the kernel bit stream in provisioning packets and transmit the packets to the accelerator sled. The provisioning packets may include additional data that direct the accelerator sled to configure the target accelerator device with the kernel.
In block 2124, the orchestrator server 1620 maps the loaded kernels according to the topology. In particular, the orchestrator server 1620 may propagate the kernel topology data to accelerator sleds in the system 1610. The accelerator sled may configure network connections between kernels based on the information provided by the kernel topology data. For example, if a kernel A executing in an accelerator device A is to communicate with a kernel B executing in an accelerator device B, the accelerator sled of kernel A may configure accelerator device A with routing information (e.g., the network address of accelerator device B) to allow kernel A to communicate with kernel B.
Referring now to FIG. 22, the orchestrator server 1620, in operation, may execute a method 2200 to scale a configuration of accelerated kernels in the system as a function of a kernel topology. As shown, the method 2200 begins in block 2202, in which the orchestrator server 1620 monitors kernel resource utilization. In particular, in block 2204, the orchestrator server 1620 collects telemetry data associated with the accelerator sleds (e.g., accelerator sleds 1640, 1650, or 1660). For example, each accelerator sled may transmit telemetry data to the orchestrator server 1620. As another example, the orchestrator server 1620 may obtain telemetry data in network devices or components (e.g., switches, NICs, etc.) in the system 1610 connecting kernels with one another.
In block 2206, the orchestrator server 1620 evaluates the monitored kernel resource utilization relative to scaling policies. More particularly, the orchestrator server 1620 determines, as a function of the scaling policies, whether the kernels presently provisioned with the accelerator devices should be scaled up or down based on the present resource utilization. As stated, the scaling policies provide one or more conditions, which if triggered, cause the orchestrator server 1620 to re-provision one or more kernels. In block 2208, the orchestrator server 1620 determines whether a scaling condition is triggered. If not, the method 2200 loops back to block 2202, in which the orchestrator server 1620 continues to monitor kernel resource utilization. Otherwise, in block 2210, the orchestrator server 1620 determines whether the triggered condition relates to power management, such as power consumption or a present power state of a given accelerator sled or device. If so, the orchestrator server 1620 may initiate kernel scaling to shift power consumption to multiple accelerator sleds, further described relative to FIGS. 23, 24A, and 24B.
If the condition is not related to power management, then in block 2212, the orchestrator server 1620 identifies one or more kernels to re-provision as a function of the triggered scaling condition. For example, a policy may specify that if latency is high in kernel A when communicating with kernel C at all accelerator devices configured with kernel C, then instances of kernel C should be scaled up. In block 2214, the orchestrator server 1620 re-provisions kernels according to the triggered scaling condition. In block 2216, the orchestrator server 1620 loads or removes the kernel instance in or from accelerator sleds based on the triggered condition. Continuing the previous example, the triggered scaling condition indicates that additional kernel C instances should be provisioned. The scaling condition may also indicate other rules for provisioning the instances, such as relative proximity to an accelerator device configured with kernel A. The orchestrator server 1620 may identify available accelerator devices on which to configure the kernel. Once identified, the orchestrator server 1620 may send a bit stream corresponding to the kernel to the accelerator devices. In block 2218, the orchestrator server 1620 updates the kernel topology data to include the additional instances and the device locations of the instances.
Referring now to FIG. 23, the orchestrator server 1620, in operation, may execute a method 2300 to provision additional instances of a kernel executing on an accelerator device as a function of a power state associated with the accelerator device. As shown, the method 2300 begins in block 2302, in which the orchestrator server 1620 detects, as a function of monitored telemetry data, that an accelerator device configured with a kernel accelerating functions for a workload is in a critical power state. The telemetry data collected from an accelerator device may include power consumption as a metric for evaluation by the orchestrator server 1620. If the power consumption exceeds a threshold specified by a scaling policy, the orchestrator server 1620 may determine that the accelerator device (or sled) is currently in a critical power state.
In block 2304, the orchestrator server 1620 identifies a kernel that is configured with the accelerator device that is in a critical power state. For example, to do so, the orchestrator server 1620 may evaluate kernel topology data to determine a configuration of the accelerator device. The configuration may specify one or more kernels presently provisioned to the accelerator device. In block 2306, the orchestrator server 1620 determines available accelerator devices in the system 1610 that can be provisioned with the kernel. For instance, the orchestrator server 1620 may evaluate accelerator devices that do not have a kernel presently configured therewith. Further, the orchestrator server 1620 may evaluate characteristics of the accelerator devices, such as device location and a present load on the accelerator device. The orchestrator server 1620, as a function of the characteristics, may select one or more of the accelerator devices on which to provision the kernel.
In block 2308, the orchestrator server 1620 provisions the identified kernel on the selected accelerator devices. In particular, in block 2310, the orchestrator server 1620 loads the kernel on the accelerator devices. To do so, the orchestrator server 1620 may retrieve a bit stream corresponding to the kernel from a predetermined data store and send the bit stream to the accelerator devices, using similar techniques to those described above. In turn, the receiving accelerator devices may configure the kernel bit stream. In block 2312, the orchestrator server 1620 updates the kernel topology data to include the provisioning of the kernel instances on the additional accelerator devices. Further, the orchestrator server 1620 propagates the updated kernel topology data to the accelerator sleds in the system 1610. As a result, the workload data processed by the kernel is shared among accelerator devices, thereby shifting power consumption.
Referring now to FIGS. 24A and 24B, diagrams of an example embodiment of provisioning additional instances of a kernel to reduce power consumption in an accelerator sled is shown. FIG. 24A depicts a kernel A 2405 and a kernel B 2410 communicating via a network 2415. The network 2415 is representative of one or more networking devices (e.g., a switch device, NIC, etc.) interconnecting the kernel A 2405 and kernel B 2410 according to a kernel topology. The kernel A 2405 accelerates a function of a workload and transmits resulting data from the function to the kernel B 2410 over the network 2415. Illustratively, the entirety of the data is transmitted to the kernel B 2410 (represented by the indication of 100% bandwidth). Further, FIG. 24A depicts the kernel B 2410 (e.g., the accelerator device associated with the kernel B) in a critical power state. As stated, telemetry data collected from the accelerator device configured with kernel B may provide a power consumption metric that may indicate the critical power state.
FIG. 24B depicts the result of scaling an additional kernel B instance (kernel B 2412) to offset power consumption to another accelerator device. The kernel B 2412 may reside an accelerator device of another sled. Illustratively, data transmitted over the network 2415 by kernel A 2405 is shared between the kernel B 2410 and kernel B 2412, represented by the “normal power state” indication in each.
Referring now to FIG. 25, the accelerator sled 1800 (or any of the accelerator sleds 1640, 1650, 1660) may, in operation, execute a method 2500 for provisioning, via an in-band communication by an accelerator device in the sled 1800, a target kernel on another accelerator device. As shown, the method 2500 begins in block 2502, where the accelerator sled 1900 pre-emptively stores, via the workload, a bit stream associated with the target kernel. For example, during provisioning of the accelerator devices according to the kernel topology, the orchestrator server 1620 may transmit the bit stream of the target kernel to the accelerator sled 1800. The accelerator sled 1800 may store the bit stream in a local data store.
In block 2504, the accelerator sled 1800 identifies one or more unused accelerator devices on which to provision the kernel. In particular, in block 2506, the kernel broadcasts a resource request packet via the kernel-to-kernel communication network. The packet is indicative of a request for accelerator devices that presently are not configured with a kernel or are otherwise not in use. The accelerator devices may reside on another sled separate from the sled 1800. In block 2506, the kernel receives a response to the request from one or more of the unused accelerator devices. The response may include data indicative of one or more characteristics of the responding accelerator device. In block 2510, the accelerator sled 1800 selects one or more of the accelerator devices as a function of the characteristics of each sled.
In block 2512, the accelerator sled 1800 configures the selected accelerator device (or devices) with the target kernel. In particular, in block 2514, the accelerator sled 1800 encapsulates the previously stored bit stream data associated with the target kernel in one or more provisioning packets. The provisioning packets may also include data causing the receiving accelerator sled to configure the bit stream data in a slot of the accelerator device. In block 2516, the accelerator sled 1800 sends the provisioning packets to the identified accelerator devices over the kernel-to-kernel network. More particularly, the accelerator sled 1800 may specify a network address of the accelerator sled hosting the accelerator device in the destination address of each packet. In turn, the target accelerator sled receives the packets and may configure the bit stream data in the accelerator device.
In block 2518, the accelerator sled 1800 optionally notifies the orchestrator server 1620 of the in-band provisioning of the target kernel on the accelerator device. For instance, the accelerator sled 1800 may transmit data indicative of the configuration, the type of the kernel, the target accelerator sled and device, and the like. In turn, the orchestrator server 1620 may update kernel topology data to include the configuration of the kernel on the accelerator device. The orchestrator server 1620 may also propagate the update to other accelerator sleds in the system 1610.
In block 2520, as a result of the in-band provisioning, the kernel configured in the accelerator sled 1800 may transmit workload data to the target kernel in the newly-provisioned accelerator device. Referring now to FIGS. 26A and 26B, diagrams of an example embodiment of provisioning, via in-band communication by an accelerator device, a kernel on another accelerator device is shown. FIG. 26A depicts a kernel A 2605 receiving processing workload data from an application 2607. For example, the application 2607 may be representative of the application 1634. The kernel A 2605 is configured to output resulting workload data to kernel B 2610 via a kernel-to-kernel communication network 2615. The network 2615 is representative of one or more networking devices (e.g., a switch device, NIC, etc.) interconnecting the kernel A 2610 and the kernel B 2610 according to a kernel topology.
The kernel A may, via the application 2607, determine that additional instances of kernel B 2610 should be provisioned on another accelerator device. For example, the application 2607 may determine that a number of connections (or other observed telemetry data) made with the kernel B exceeds some specified threshold. In some cases, it is preferable for the kernel A 2605 to provision an additional instance of kernel B on another accelerator device. For instance, the accelerator sled configured with kernel A may detect an increasing load on resource utilization on kernel B sooner than the orchestrator server 1620 does. Further, in some cases, provisioning an additional kernel B 2610 instance may have a lower operational cost, such as in network latency, than if the orchestrator server 1620 provisions the kernel (e.g., in cases where kernel A provisions an additional kernel B 2610 instance on the same sled as kernel A).
Illustratively, the kernel A 2605 broadcasts a resource request packet via the kernel-to-kernel communication network 2615 (as indicated by arrow 2618). The resource request packet may be received at unused resources 2612, which are representative of accelerator devices in the system 1610 that are not presently configured with a kernel instance or devices that include available slots. The accelerator devices may reside on the same accelerator sled as kernel A 2605 or on a separate accelerator sled. Unused resources 2612 that are available to be configured with an instance of kernel B may send a response to kernel A 2605, which may include data indicative of characteristics of an unused resource 2612. The kernel A 2605 may, via the application 2607, determine one of the unused resources 2612 to provision with the kernel B instance as a function of the characteristics.
FIG. 26B depicts kernel A 2605 provisioning a new kernel B instance (kernel B 2610′). In particular, the kernel A 2605 sends one or more provisioning packets including a bit stream corresponding to the determined unused resource 2612 (e.g., an accelerator device) over the network 2615 (as indicated by arrow 2620). Once received, the accelerator device may provision the bit stream in one of the slots therein. Further, the kernel A 2605 may send an update to the kernel topology data to the orchestrator server 1620. As a result, the kernel A 2605 may output kernel data to both instances of kernel B 2610 and 2610′.

EXAMPLES

Illustrative examples of the technologies disclosed herein are provided below. An embodiment of the technologies may include any one or more, and any combination of, the examples described below.
Example 1 includes a compute device, comprising a compute engine to (i) receive, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled, (ii) determine, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels, and (iii) configure the plurality of kernels on the plurality of accelerator devices according to the determined topology.
Example 2 includes the subject matter of Example 1, and wherein to determine the topology of the plurality of kernels comprises to determine one or more characteristics of each of the accelerator devices; and determine one or more configurations of the plurality of kernels on the plurality of accelerator devices.
Example 3 includes the subject matter of any of Examples 1 and 2, and wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.
Example 4 includes the subject matter of any of Examples 1-3, and wherein to determine the topology comprises to determine a flow of output data from the first kernel to be used as input data to the second kernel.
Example 5 includes the subject matter of any of Examples 1-4, and wherein to configure the plurality of kernels comprises to identify one of the plurality of accelerator devices to provision with one of the plurality of kernels; retrieve a bit stream associated with the one of the plurality of kernels; load the bit stream onto the one of the plurality of accelerator devices; and map the one of the plurality of kernels to another of the plurality of kernels according to the topology.
Example 6 includes the subject matter of any of Examples 1-5, and wherein the compute engine is further to monitor resource utilization in each of the plurality of kernels; evaluate the resource utilization relative to one or more scaling policies; and determine, based on the evaluation, whether a condition to scale one of the plurality of kernels has been triggered.
Example 7 includes the subject matter of any of Examples 1-6, and wherein to monitor resource utilization in each of the plurality of kernels comprises to collect telemetry data reported by each of the plurality of accelerator sleds, wherein the telemetry data is indicative of the resource utilization in each of the plurality of accelerator sleds during execution of the workload.
Example 8 includes the subject matter of any of Examples 1-7, and wherein the compute engine is further to identify, in response to a determination that a condition to scale one of the plurality of kernels has been triggered, one of the kernels to re-provision as a function of the triggered condition; re-provision the one of the kernels according to the scaling policy; and update the kernel topology as a function of the re-provision.
Example 9 includes the subject matter of any of Examples 1-8, and wherein to re-provision the one of the kernels according to the scaling policy comprises to provision the one of the kernels on an additional accelerator device of the plurality of accelerator devices.
Example 10 includes the subject matter of any of Examples 1-9, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises to identify a first accelerator device on which the one of the kernels is provisioned, wherein the first accelerator device resides on a first accelerator sled.
Example 11 includes the subject matter of any of Examples 1-10, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on a second accelerator sled.
Example 12 includes the subject matter of any of Examples 1-11, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on the first accelerator sled.
Example 13 includes the subject matter of any of Examples 1-12, and wherein to re-provision the one of the kernels according to the scaling policy comprises to identify a first accelerator device on which the one of the kernels is provisioned; and remove the one of the kernels from the first accelerator device.
Example 14 includes a method comprising receiving, by a compute device and from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled; determining, by the compute device and as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels; and configuring, by the compute device, the plurality of kernels on the plurality of accelerator devices according to the determined topology.
Example 15 includes the subject matter of Example 14, and wherein determining the topology of the plurality of kernels comprises determining one or more characteristics of each of the accelerator devices; and determining one or more configurations of the plurality of kernels on the plurality of accelerator devices.
Example 16 includes the subject matter of any of Examples 14 and 15, and wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.
Example 17 includes the subject matter of any of Examples 14-16, and wherein determining the topology comprises determining a flow of output data from the first kernel to be used as input data to the second kernel.
Example 18 includes the subject matter of any of Examples 14-17, and wherein configuring the plurality of kernels comprises identifying one of the plurality of accelerator devices to provision with one of the plurality of kernels; retrieving a bit stream associated with the one of the plurality of kernels; loading the bit stream onto the one of the plurality of accelerator devices; and mapping the one of the plurality of kernels to another of the plurality of kernels according to the topology.
Example 19 includes the subject matter of any of Examples 14-18, and further including monitoring, by the compute device, resource utilization in each of the plurality of kernels; evaluating, by the compute device, the resource utilization relative to one or more scaling policies; and determining, by the compute device and based on the evaluation, whether a condition to scale one of the plurality of kernels has been triggered.
Example 20 includes the subject matter of any of Examples 14-19, and wherein monitoring resource utilization in each of the plurality of kernels comprises collecting, by the compute device, telemetry data reported by each of the plurality of accelerator sleds, wherein the telemetry data is indicative of the resource utilization in each of the plurality of accelerator sleds during execution of the workload.
Example 21 includes the subject matter of any of Examples 14-20, and further including identifying, by the compute device and in response to a determination that a condition to scale one of the plurality of kernels has been triggered, one of the kernels to re-provision as a function of the triggered condition; re-provisioning, by the compute device, the one of the kernels according to the scaling policy; and updating, by the compute device, the kernel topology as a function of the re-provisioning.
Example 22 includes the subject matter of any of Examples 14-21, and wherein re-provisioning the one of the kernels according to the scaling policy comprises provisioning the one of the kernels on an additional accelerator device of the plurality of accelerator devices.
Example 23 includes the subject matter of any of Examples 14-22, and wherein provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises identifying a first accelerator device on which the one of the kernels is provisioned, wherein the first accelerator device resides on a first accelerator sled.
Example 24 includes the subject matter of any of Examples 14-23, and wherein provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises provisioning the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on a second accelerator sled.
Example 25 includes the subject matter of any of Examples 14-24, and wherein provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises provisioning the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on the first accelerator sled.
Example 26 includes the subject matter of any of Examples 14-25, and wherein re-provisioning the one of the kernels according to the scaling policy comprises identifying a first accelerator device on which the one of the kernels is provisioned; and removing the one of the kernels from the first accelerator device.
Example 27 includes one or more machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a compute device to perform the method of any of Examples 14-26.
Example 28 includes a compute device comprising means for performing the method of any of Examples 14-26.
Example 29 includes a compute device comprising a compute engine to perform the method of any of Examples 14-26.
Example 30 includes a compute device comprising topology manager circuitry to receive, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled, determine, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels, and configure the plurality of kernels on the plurality of accelerator devices according to the determined topology.
Example 31 includes the subject matter of Example 30, and wherein to determine the topology of the plurality of kernels comprises to determine one or more characteristics of each of the accelerator devices; and determine one or more configurations of the plurality of kernels on the plurality of accelerator devices.
Example 32 includes the subject matter of any of Examples 30 and 31, and wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.
Example 33 includes the subject matter of any of Examples 30-32, and wherein to determine the topology comprises to determine a flow of output data from the first kernel to be used as input data to the second kernel.
Example 34 includes the subject matter of any of Examples 30-33, and wherein to configure the plurality of kernels comprises to identify one of the plurality of accelerator devices to provision with one of the plurality of kernels; retrieve a bit stream associated with the one of the plurality of kernels; load the bit stream onto the one of the plurality of accelerator devices; and map the one of the plurality of kernels to another of the plurality of kernels according to the topology.
Example 35 includes the subject matter of any of Examples 30-34, and wherein the topology manager circuitry is further to monitor resource utilization in each of the plurality of kernels; evaluate the resource utilization relative to one or more scaling policies; and determine, based on the evaluation, whether a condition to scale one of the plurality of kernels has been triggered.
Example 36 includes the subject matter of any of Examples 30-35, and wherein to monitor resource utilization in each of the plurality of kernels comprises to collect telemetry data reported by each of the plurality of accelerator sleds, wherein the telemetry data is indicative of the resource utilization in each of the plurality of accelerator sleds during execution of the workload.
Example 37 includes the subject matter of any of Examples 30-36, and wherein the topology manager circuitry is further to identify, in response to a determination that a condition to scale one of the plurality of kernels has been triggered, one of the kernels to re-provision as a function of the triggered condition; re-provision the one of the kernels according to the scaling policy; and update the kernel topology as a function of the re-provision.
Example 38 includes the subject matter of any of Examples 30-37, and wherein to re-provision the one of the kernels according to the scaling policy comprises to provision the one of the kernels on an additional accelerator device of the plurality of accelerator devices.
Example 39 includes the subject matter of any of Examples 30-38, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises to identify a first accelerator device on which the one of the kernels is provisioned, wherein the first accelerator device resides on a first accelerator sled.
Example 40 includes the subject matter of any of Examples 30-39, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on a second accelerator sled.
Example 41 includes the subject matter of any of Examples 30-40, and wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on the first accelerator sled.
Example 42 includes the subject matter of any of Examples 30-41, and wherein to re-provision the one of the kernels according to the scaling policy comprises to identify a first accelerator device on which the one of the kernels is provisioned; and remove the one of the kernels from the first accelerator device.
Example 43 includes a compute device comprising circuitry for receiving, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled; means for determining, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels; and means for configuring the plurality of kernels on the plurality of accelerator devices according to the determined topology.
Example 44 includes the subject matter of Example 43, and wherein the means for determining the topology of the plurality of kernels comprises circuitry for determining one or more characteristics of each of the accelerator devices; and circuitry for determining one or more configurations of the plurality of kernels on the plurality of accelerator devices.
Example 45 includes the subject matter of any of Examples 43 and 44, and wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.
Example 46 includes the subject matter of any of Examples 43-45, and wherein the means for determining the topology comprises circuitry for determining a flow of output data from the first kernel to be used as input data to the second kernel.
Example 47 includes the subject matter of any of Examples 43-46, and wherein the means for configuring the plurality of kernels comprises circuitry for identifying one of the plurality of accelerator devices to provision with one of the plurality of kernels; circuitry for retrieving a bit stream associated with the one of the plurality of kernels; circuitry for loading the bit stream onto the one of the plurality of accelerator devices; and circuitry for mapping the one of the plurality of kernels to another of the plurality of kernels according to the topology.
Example 48 includes the subject matter of any of Examples 43-47, and further including means for monitoring resource utilization in each of the plurality of kernels; means for evaluating the resource utilization relative to one or more scaling policies; and means for determining, based on the evaluation, whether a condition to scale one of the plurality of kernels has been triggered.
Example 49 includes the subject matter of any of Examples 43-48, and wherein the means for monitoring resource utilization in each of the plurality of kernels comprises circuitry for collecting telemetry data reported by each of the plurality of accelerator sleds, wherein the telemetry data is indicative of the resource utilization in each of the plurality of accelerator sleds during execution of the workload.
Example 50 includes the subject matter of any of Examples 43-49, and further including means for identifying, in response to a determination that a condition to scale one of the plurality of kernels has been triggered, one of the kernels to re-provision as a function of the triggered condition; means for re-provisioning the one of the kernels according to the scaling policy; and means for updating the kernel topology as a function of the re-provisioning.
Example 51 includes the subject matter of any of Examples 43-50, and wherein the means for re-provisioning the one of the kernels according to the scaling policy comprises circuitry for provisioning the one of the kernels on an additional accelerator device of the plurality of accelerator devices.
Example 52 includes the subject matter of any of Examples 43-51, and wherein the circuitry for provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises circuitry for identifying a first accelerator device on which the one of the kernels is provisioned, wherein the first accelerator device resides on a first accelerator sled.
Example 53 includes the subject matter of any of Examples 43-52, and wherein the circuitry for provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises circuitry for provisioning the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on a second accelerator sled.
Example 54 includes the subject matter of any of Examples 43-53, and wherein the circuitry for provisioning the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises circuitry for provisioning the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on the first accelerator sled.
Example 55 includes the subject matter of any of Examples 43-54, and wherein the means for re-provisioning the one of the kernels according to the scaling policy comprises circuitry for identifying a first accelerator device on which the one of the kernels is provisioned; and circuitry for removing the one of the kernels from the first accelerator device.
Example 56 includes a compute device comprising a compute engine to (i) detect whether a first accelerator device of a plurality of accelerator devices is in a critical power state, wherein the first accelerator device is configured with a kernel of a plurality of kernels and the critical power state is indicative that a present power usage of the first accelerator device satisfies a predefined threshold, (ii) determine a second accelerator device of the plurality of accelerator devices that is available to be configured with the kernel, and (iii) configure, in response to a determination that the first accelerator device is in a critical power state, the second accelerator device with the kernel, wherein the kernel configured with the first accelerator device is to share acceleration of the execution of a workload with the kernel configured with the second accelerator device.
Example 57 includes the subject matter of Example 56, and wherein the first accelerator device resides on a first accelerator sled, and wherein the second accelerator device resides on a second accelerator sled.
Example 58 includes the subject matter of any of Examples 56 and 57, and wherein the first accelerator sled and the second accelerator sled are located on one of a plurality of racks in a data center.
Example 59 includes the subject matter of any of Examples 56-58, and wherein to determine the second accelerator device of the plurality of accelerator devices comprises to identify one or more of the accelerator devices that are not in a critical power state.
Example 60 includes the subject matter of any of Examples 56-59, and wherein the first accelerator device is configured with the kernel according to a kernel topology, wherein the kernel topology maps data communication between the kernel and a second kernel.
Example 61 includes the subject matter of any of Examples 56-60, and wherein the compute engine is further to update the kernel topology based on the configuration of the second accelerator device with the kernel.
Example 62 includes the subject matter of any of Examples 56-61, and wherein to detect whether a first accelerator device of a plurality of accelerator devices is in a power critical state comprises to monitor telemetry data in the first accelerator device, wherein the telemetry data is indicative of a present power state in the accelerator device.
Example 63 includes the subject matter of any of Examples 56-62, and wherein the compute engine is further to configure the first accelerator device and the second accelerator device configured with the kernel to receive output data from another kernel of the plurality of kernels.
Example 64 includes a method comprising detecting whether a first accelerator device of a plurality of accelerator devices is in a critical power state, wherein the first accelerator device is configured with a kernel of a plurality of kernels and the critical power state is indicative that a present power usage of the first accelerator device satisfies a predefined threshold; determining a second accelerator device of the plurality of accelerator devices that is available to be configured with the kernel; configuring, in response to a determination that the first accelerator device is in a critical power state, the second accelerator device with the kernel, wherein the kernel configured with the first accelerator device is to share acceleration of the execution of a workload with the kernel configured with the second accelerator device.
Example 65 includes the subject matter of Example 64, and wherein the first accelerator device resides on a first accelerator sled, and wherein the second accelerator device resides on a second accelerator sled.
Example 66 includes the subject matter of any of Examples 64 and 65, and wherein the first accelerator sled and the second accelerator sled are located on one of a plurality of racks in a data center.
Example 67 includes the subject matter of any of Examples 64-66, and wherein determining the second accelerator device of the plurality of accelerator devices comprises identifying one or more of the accelerator devices that are not in a critical power state.
Example 68 includes the subject matter of any of Examples 64-67, and wherein the first accelerator device is configured with the kernel according to a kernel topology, wherein the kernel topology maps data communication between the kernel and a second kernel.
Example 69 includes the subject matter of any of Examples 64-68, and further including updating the kernel topology based on the configuration of the second accelerator device with the kernel.
Example 70 includes the subject matter of any of Examples 64-69, and wherein detecting whether a first accelerator device of a plurality of accelerator devices is in a power critical state comprises monitoring telemetry data in the first accelerator device, wherein the telemetry data is indicative of a present power state in the accelerator device.
Example 71 includes the subject matter of any of Examples 64-70, and further including configuring the first accelerator device and the second accelerator device configured with the kernel to receive output data from another kernel of the plurality of kernels.
Example 72 includes one or more machine-readable storage media comprising a plurality of instructions stored therein that, in response to being executed, cause a compute device to perform the method of any of Examples 64-71.
Example 73 includes a compute device comprising means for performing the method of any of Examples 64-71.
Example 74 includes a compute device comprising a compute engine to perform the method of any of Examples 64-71.
Example 75 includes a compute device comprising topology manager circuitry to (i) detect whether a first accelerator device of a plurality of accelerator devices is in a critical power state, wherein the first accelerator device is configured with a kernel of a plurality of kernels and the critical power state is indicative that a present power usage of the first accelerator device satisfies a predefined threshold, (ii) determine a second accelerator device of the plurality of accelerator devices that is available to be configured with the kernel, and (iii) configure, in response to a determination that the first accelerator device is in a critical power state, the second accelerator device with the kernel, wherein the kernel configured with the first accelerator device is to share acceleration of the execution of a workload with the kernel configured with the second accelerator device.
Example 76 includes the subject matter of Example 75, and wherein the first accelerator device resides on a first accelerator sled, and wherein the second accelerator device resides on a second accelerator sled.
Example 77 includes the subject matter of any of Examples 75 and 76, and wherein the first accelerator sled and the second accelerator sled are located on one of a plurality of racks in a data center.
Example 78 includes the subject matter of any of Examples 75-77, and wherein to determine the second accelerator device of the plurality of accelerator devices comprises to identify one or more of the accelerator devices that are not in a critical power state.
Example 79 includes the subject matter of any of Examples 75-78, and wherein the first accelerator device is configured with the kernel according to a kernel topology, wherein the kernel topology maps data communication between the kernel and a second kernel.
Example 80 includes the subject matter of any of Examples 75-79, and wherein the topology manager circuitry is further to update the kernel topology based on the configuration of the second accelerator device with the kernel.
Example 81 includes the subject matter of any of Examples 75-80, and wherein to detect whether a first accelerator device of a plurality of accelerator devices is in a power critical state comprises to monitor telemetry data in the first accelerator device, wherein the telemetry data is indicative of a present power state in the accelerator device.
Example 82 includes the subject matter of any of Examples 75-81, and wherein the topology manager circuitry is further to configure the first accelerator device and the second accelerator device configured with the kernel to receive output data from another kernel of the plurality of kernels.
Example 83 includes a compute device comprising circuitry for detecting whether a first accelerator device of a plurality of accelerator devices is in a critical power state, wherein the first accelerator device is configured with a kernel of a plurality of kernels and the critical power state is indicative that a present power usage of the first accelerator device satisfies a predefined threshold; means for determining a second accelerator device of the plurality of accelerator devices that is available to be configured with the kernel; means for configuring, in response to a determination that the first accelerator device is in a critical power state, the second accelerator device with the kernel, wherein the kernel configured with the first accelerator device is to share acceleration of the execution of a workload with the kernel configured with the second accelerator device.
Example 84 includes the subject matter of Example 83, and wherein the first accelerator device resides on a first accelerator sled, and wherein the second accelerator device resides on a second accelerator sled.
Example 85 includes the subject matter of any of Examples 83 and 84, and wherein the first accelerator sled and the second accelerator sled are located on one of a plurality of racks in a data center.
Example 86 includes the subject matter of any of Examples 83-85, and wherein the means for determining the second accelerator device of the plurality of accelerator devices comprises circuitry for identifying one or more of the accelerator devices that are not in a critical power state.
Example 87 includes the subject matter of any of Examples 83-86, and wherein the first accelerator device is configured with the kernel according to a kernel topology, wherein the kernel topology maps data communication between the kernel and a second kernel.
Example 88 includes the subject matter of any of Examples 83-87, and further including means for updating the kernel topology based on the configuration of the second accelerator device with the kernel.
Example 89 includes the subject matter of any of Examples 83-88, and wherein the circuitry for detecting whether a first accelerator device of a plurality of accelerator devices is in a power critical state comprises circuitry for monitoring telemetry data in the first accelerator device, wherein the telemetry data is indicative of a present power state in the accelerator device.
Example 90 includes the subject matter of any of Examples 83-89, and further including means for configuring the first accelerator device and the second accelerator device configured with the kernel to receive output data from another kernel of the plurality of kernels.
Example 91 includes a sled comprising a compute engine to (i) receive a request by a workload to provision a first kernel on one of a plurality of accelerator devices, wherein the first kernel defines an accelerator device configuration that is usable by one of the plurality of accelerator devices to perform one or more accelerated functions associated with the workload, (ii) identify a first accelerator device of the plurality of accelerator devices on which to provision the first kernel, and (iii) configure the first accelerator device with the first kernel.
Example 92 includes the subject matter of Example 91, and wherein a second accelerator device in the plurality of accelerator devices is provisioned with a second kernel, wherein the second accelerator device resides on the sled.
Example 93 includes the subject matter of any of Examples 91 and 92, and wherein the first accelerator device resides on an accelerator sled different from the sled.
Example 94 includes the subject matter of any of Examples 91-93, and wherein to identify the first accelerator device of the plurality of accelerator devices comprises to broadcast a packet to each of the plurality of accelerator devices, wherein the packet is indicative of a request; and receive a response to the request from at least one of the plurality of accelerator devices including the first accelerator device.
Example 95 includes the subject matter of any of Examples 91-94, and wherein to identify the first accelerator device of the plurality of accelerator devices further comprises to select the first accelerator device from the at least one of the plurality of accelerator devices as a function of one or more characteristics of each of the at least one of the plurality of accelerator devices.
Example 96 includes the subject matter of any of Examples 91-95, and wherein the one or more characteristics comprises at least one of an availability, a power state, or a location of the sled.
Example 97 includes the subject matter of any of Examples 91-96, and wherein to configure the identified first accelerator device comprises to encapsulate bit stream data associated with the first kernel in one or more provisioning packets; and send the one or more provisioning packets to the first accelerator device.
Example 98 includes the subject matter of any of Examples 91-97, and wherein the compute engine is further to notify a server of the configuration of the first accelerator device with the first kernel.
Example 99 includes a method comprising receiving, by a sled, a request by a workload to provision a first kernel on one of a plurality of accelerator devices, wherein the first kernel defines an accelerator device configuration that is usable by one of the plurality of accelerator devices to perform one or more accelerated functions associated with the workload; identifying, by the sled, a first accelerator device of the plurality of accelerator devices on which to provision the first kernel; and configuring, by the sled, the first accelerator device with the first kernel.
Example 100 includes the subject matter of Example 99, and wherein a second accelerator device in the plurality of accelerator devices is provisioned with a second kernel, wherein the second accelerator device resides on the sled.
Example 101 includes the subject matter of any of Examples 99 and 100, and wherein the first accelerator device resides on an accelerator sled different from the sled.
Example 102 includes the subject matter of any of Examples 99-101, and wherein identifying the first accelerator device of the plurality of accelerator devices comprises broadcasting a packet to each of the plurality of accelerator devices, wherein the packet is indicative of a request; and receiving a response to the request from at least one of the plurality of accelerator devices including the first accelerator device.
Example 103 includes the subject matter of any of Examples 99-102, and wherein identifying the first accelerator device of the plurality of accelerator devices further comprises selecting the first accelerator device from the at least one of the plurality of accelerator devices as a function of one or more characteristics of each of the at least one of the plurality of accelerator devices.
Example 104 includes the subject matter of any of Examples 99-103, and wherein the one or more characteristics comprises at least one of an availability, a power state, or a location of the sled.
Example 105 includes the subject matter of any of Examples 99-104, and wherein configuring the identified first accelerator device comprises encapsulating bit stream data associated with the first kernel in one or more provisioning packets; and sending the one or more provisioning packets to the first accelerator device.
Example 106 includes the subject matter of any of Examples 99-105, and further including notifying, by the sled, a server of the configuration of the first accelerator device with the first kernel.
Example 107 includes one or more machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a sled to perform the method of any of Examples 99-106.
Example 108 includes a sled comprising means for performing the method of any of Examples 99-106.
Example 109 includes a sled comprising a compute engine to perform the method of any of Examples 99-106.
Example 110 includes a sled comprising accelerator device manager circuitry to (i) receive a request by a workload to provision a first kernel on one of a plurality of accelerator devices, wherein the first kernel defines an accelerator device configuration that is usable by one of the plurality of accelerator devices to perform one or more accelerated functions associated with the workload, (ii) identify a first accelerator device of the plurality of accelerator devices on which to provision the first kernel, and (iii) configure the first accelerator device with the first kernel.
Example 111 includes the subject matter of Example 110, and wherein a second accelerator device in the plurality of accelerator devices is provisioned with a second kernel, wherein the second accelerator device resides on the sled.
Example 112 includes the subject matter of any of Examples 110 and 111, and wherein the first accelerator device resides on an accelerator sled different from the sled.
Example 113 includes the subject matter of any of Examples 110-112, and wherein to identify the first accelerator device of the plurality of accelerator devices comprises to broadcast a packet to each of the plurality of accelerator devices, wherein the packet is indicative of a request; and receive a response to the request from at least one of the plurality of accelerator devices including the first accelerator device.
Example 114 includes the subject matter of any of Examples 110-113, and wherein to identify the first accelerator device of the plurality of accelerator devices further comprises to select the first accelerator device from the at least one of the plurality of accelerator devices as a function of one or more characteristics of each of the at least one of the plurality of accelerator devices.
Example 115 includes the subject matter of any of Examples 110-114, and wherein the one or more characteristics comprises at least one of an availability, a power state, or a location of the sled.
Example 116 includes the subject matter of any of Examples 110-115, and wherein to configure the identified first accelerator device comprises to encapsulate bit stream data associated with the first kernel in one or more provisioning packets; and send the one or more provisioning packets to the first accelerator device.
Example 117 includes the subject matter of any of Examples 110-116, and wherein the accelerator device manager circuitry is further to notify a server of the configuration of the first accelerator device with the first kernel.
Example 118 includes a sled comprising circuitry for receiving a request by a workload to provision a first kernel on one of a plurality of accelerator devices, wherein the first kernel defines an accelerator device configuration that is usable by one of the plurality of accelerator devices to perform one or more accelerated functions associated with the workload; means for identifying a first accelerator device of the plurality of accelerator devices on which to provision the first kernel; and means for configuring the first accelerator device with the first kernel.
Example 119 includes the subject matter of Example 118, and wherein a second accelerator device in the plurality of accelerator devices is provisioned with a second kernel, wherein the second accelerator device resides on the sled.
Example 120 includes the subject matter of any of Examples 118 and 119, and wherein the first accelerator device resides on an accelerator sled different from the sled.
Example 121 includes the subject matter of any of Examples 118-120, and wherein the means for identifying the first accelerator device of the plurality of accelerator devices comprises circuitry for broadcasting a packet to each of the plurality of accelerator devices, wherein the packet is indicative of a request; and circuitry for receiving a response to the request from at least one of the plurality of accelerator devices including the first accelerator device.
Example 122 includes the subject matter of any of Examples 118-121, and wherein the means for identifying the first accelerator device of the plurality of accelerator devices further comprises circuitry for selecting the first accelerator device from the at least one of the plurality of accelerator devices as a function of one or more characteristics of each of the at least one of the plurality of accelerator devices.
Example 123 includes the subject matter of any of Examples 118-122, and wherein the one or more characteristics comprises at least one of an availability, a power state, or a location of the sled.
Example 124 includes the subject matter of any of Examples 118-123, and wherein the means for configuring the identified first accelerator device comprises circuitry for encapsulating bit stream data associated with the first kernel in one or more provisioning packets; and circuitry for sending the one or more provisioning packets to the first accelerator device.
Example 125 includes the subject matter of any of Examples 118-124, and further including circuitry for notifying a server of the configuration of the first accelerator device with the first kernel.

Claims

1. A compute device, comprising:

a compute engine to (i) receive, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled, (ii) determine, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels, and (iii) configure the plurality of kernels on the plurality of accelerator devices according to the determined topology.

2. The compute device of claim 1, wherein to determine the topology of the plurality of kernels comprises to:

determine one or more characteristics of each of the accelerator devices; and

determine one or more configurations of the plurality of kernels on the plurality of accelerator devices.

3. The compute device of claim 2, wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.

4. The compute device of claim 1, wherein to determine the topology comprises to determine a flow of output data from the first kernel to be used as input data to the second kernel.

5. The compute device of claim 1, wherein to configure the plurality of kernels comprises to:

identify one of the plurality of accelerator devices to provision with one of the plurality of kernels;

retrieve a bit stream associated with the one of the plurality of kernels;

load the bit stream onto the one of the plurality of accelerator devices; and

map the one of the plurality of kernels to another of the plurality of kernels according to the topology.

6. The compute device of claim 1, wherein the compute engine is further to:

monitor resource utilization in each of the plurality of kernels;

evaluate the resource utilization relative to one or more scaling policies; and

determine, based on the evaluation, whether a condition to scale one of the plurality of kernels has been triggered.

7. The compute device of claim 6, wherein to monitor resource utilization in each of the plurality of kernels comprises to:

collect telemetry data reported by each of the plurality of accelerator sleds, wherein the telemetry data is indicative of the resource utilization in each of the plurality of accelerator sleds during execution of the workload.

8. The compute device of claim 6, wherein the compute engine is further to:

identify, in response to a determination that a condition to scale one of the plurality of kernels has been triggered, one of the kernels to re-provision as a function of the triggered condition;

re-provision the one of the kernels according to the scaling policy; and

update the kernel topology as a function of the re-provision.

9. The compute device of claim 8, wherein to re-provision the one of the kernels according to the scaling policy comprises to:

provision the one of the kernels on an additional accelerator device of the plurality of accelerator devices.

10. The compute device of claim 9, wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises to:

identify a first accelerator device on which the one of the kernels is provisioned, wherein the first accelerator device resides on a first accelerator sled.

11. The compute device of claim 10, wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to:

provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on a second accelerator sled.

12. The compute device of claim 10, wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to:

provision the one of the kernels on the additional accelerator device, wherein the additional accelerator device resides on the first accelerator sled.

13. The compute device of claim 8, wherein to re-provision the one of the kernels according to the scaling policy comprises to:

identify a first accelerator device on which the one of the kernels is provisioned; and

remove the one of the kernels from the first accelerator device.

14. One or more machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a compute device to:

receive, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled;

determine, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels; and

configure the plurality of kernels on the plurality of accelerator devices according to the determined topology.

15. The one or more machine-readable storage media of claim 14, wherein to determine the topology of the plurality of kernels comprises to:

determine one or more characteristics of each of the accelerator devices; and

16. The one or more machine-readable storage media of claim 15, wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.

17. The one or more machine-readable storage media of claim 14, wherein to determine the topology comprises to determine a flow of output data from the first kernel to be used as input data to the second kernel.

18. The one or more machine-readable storage media of claim 14, wherein to configure the plurality of kernels comprises to:

retrieve a bit stream associated with the one of the plurality of kernels;

load the bit stream onto the one of the plurality of accelerator devices; and

19. The one or more machine-readable storage media of claim 14, wherein the plurality of instructions further cause the compute device to:

monitor resource utilization in each of the plurality of kernels;

evaluate the resource utilization relative to one or more scaling policies; and

20. The one or more machine-readable storage media of claim 19, wherein to monitor resource utilization in each of the plurality of kernels comprises to:

21. The one or more machine-readable storage media of claim 19, wherein the plurality of instructions further cause the compute device to:

re-provision the one of the kernels according to the scaling policy; and

update the kernel topology as a function of the re-provisioning.

22. The one or more machine-readable storage media of claim 21, wherein to re-provision the one of the kernels according to the scaling policy comprises to:

23. The one or more machine-readable storage media of claim 22, wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices comprises to:

24. The one or more machine-readable storage media of claim 23, wherein to provision the one of the kernels on the additional accelerator device of the plurality of accelerator devices further comprises to:

25. A compute device comprising:

circuitry for receiving, from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled;

means for determining, as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels; and

means for configuring the plurality of kernels on the plurality of accelerator devices according to the determined topology.

26. A method comprising:

receiving, by a compute device and from a sled, a kernel configuration request to provision a plurality of kernels on a plurality of accelerator devices to accelerate a task of a workload executed by the sled;

determining, by the compute device and as a function of one or more requirements of the workload, a topology of the plurality of kernels to service the kernel configuration request, wherein the topology maps data communication between a first kernel and a second kernel of the plurality of kernels; and

configuring, by the compute device, the plurality of kernels on the plurality of accelerator devices according to the determined topology.

27. The method of claim 26, wherein determining the topology of the plurality of kernels comprises:

determining one or more characteristics of each of the accelerator devices; and

determining one or more configurations of the plurality of kernels on the plurality of accelerator devices.

28. The method of claim 27, wherein the one or more characteristics includes at least one of an availability, a power state, or a device location.