KR20010006934A - Dynamic i/o allocation in a partitioned computer system - Google Patents

Abstract

PURPOSE: A computer system and a device controller are provided to operate a multiprocessor computer system, and to assign resource of the multiprocessor computer system, and to share a common I/O device set between nodes of the multiprocessor computer system. CONSTITUTION: A node(200) includes an SMP(Symmetrical MultiProcessing) processor(202) and an associative memory(204) shared with different processor. The SMP processor(202) accesses a PHB(206), and the PHB(206) accesses a USB controller(208) and a PCI slot(210). An I/O device(212) accesses the PCI slot(210), and the I/O device(212) does not share other nodes. The USB controller(208) provides a node input access about a CI/OC(Cabinet Input/Output Controller)(216), and the CI/OC(216) is controlled by a service processor(214). The CI/OC(216) accesses other nodes through a node input(218), and controls that the node inputs(218) share a device(220). At this time, the node(200) and other node(218) are accessed to the CI/OC(216) by the same input node.

Description

Computer system and device controller {DYNAMIC I / O ALLOCATION IN A PARTITIONED COMPUTER SYSTEM}

The present invention relates generally to multiprocessor systems, and more particularly to resource allocation between processors in a partitioned multiprocessor system. More specifically, the present invention relates to a method of sharing a common set of I / O devices between multiprocessor computer system nodes.

Multiprocessor computer systems are well known in the art and provide increased processing power by allowing processing tasks to be split between different system processors. In a typical system, each processor can access all of the system resources. That is, both system resources, such as memory and I / O devices, are shared among all system processors. Typically, some of the system resources may be divided between processors. For example, although each processor may access shared memory, this memory may be partitioned so that each processor has its own workspace. In a non-uniform memory access (NUMA) system, each processor has its own memory portion and can also access memory owned by another processor.

Recently, symmetric multiprocessor (SMP) systems have been partitioned to operate as multiple independent computer systems. For example, a single system with eight processors could be configured such that each of the eight processors (or multiple groups of one or more processors) is considered a separate system for processing purposes. Each such "virtual" system will have a copy of its own operating system, which can then be assigned tasks independently or otherwise operate together as a processing cluster, providing high speed processing and improved reliability.

Typically, there is also a "service" processor in a multiprocessor system, which is a startup of the overall system, including system configuration and data routing to or from a particular processor on a shared bus and device. And manage operations.

In a single multiprocessor system, when multiple virtual machines are configured to operate as a cluster, each cluster node communicates with the other nodes in the multiprocessor system to perform quorum negotiation and validation, and "heartbeats". Software support must be provided to enable other quorum functions to be performed using any cluster communication technique. When this is accomplished, if one of the processors fails-this will make the node unavailable for the cluster-the job assigned to that node will be re-located between the remaining processors (nodes) using standard cluster technology. Can be assigned.

Typically, when a multiprocessor system is divided into multiple virtual systems, each virtual system has its own copy of the operating system, and the same operating system is used for each virtual system. Since each processor runs the same operating system, providing resource allocation between processors is relatively easy.

One characteristic of large multiprocessor systems is that they are typically used in large processing tasks, so they use relatively few typical I / O such as keyboards, displays, removable media drives, and so on. Need them. However, these devices can be removed because they are sometimes needed. Making these devices available to each node in a multiprocessor system introduces unnecessary redundancy in managing and maintaining devices and the waste of providing redundant devices that are rarely used. Thus, there is a need for means to allow multiple partitions or nodes in a large multiprocessor to share a single set of I / O devices.

It is a first object of the present invention to provide a system and method for the operation of a multiprocessor computer system.

Another object of the present invention is to provide a system and method for improving resource allocation in a multiprocessor computer system.

It is yet another object of the present invention to provide a system and method for sharing a common set of I / O devices among nodes of a multiprocessor computer system.

Thus, systems and methods are provided that allow multiple nodes of a multiprocessor system to share a set of I / O devices. A cabinet input / output controller (CI / OC) is provided that manages the connection between a multiprocessor system node and a common I / O device, allowing an individual node to have exclusive access to one or more of its target devices. Each node communicates with the CI / OC through a service processor, which interconnects the USB controllers of the various I / O devices and nodes. In another embodiment, USB to the ISA bridge is also included to allow attachment of legacy I / O devices. The above objects as well as the additional objects, shapes and advantages of the present invention will become apparent from the following detailed description.

1 illustrates a multiprocessor computer system in accordance with a preferred embodiment of the present invention,

2 illustrates an SMP node connected to a CI / OC according to a preferred embodiment of the present invention.

3 shows a block diagram of CI / OC in accordance with a preferred embodiment of the present invention,

4 is a block diagram of a CI / OC and service processor according to a preferred embodiment of the present invention;

5 shows a block diagram of several I / O devices connected to a hub, downstream of CI / OC, in accordance with a preferred embodiment of the present invention.

6 illustrates a flow diagram for using a common I / O system in accordance with a preferred embodiment of the present invention.

Explanation of symbols for the main parts of the drawings

100: data processing system 101: processor

103: Level 2 Cache 106: System Bus

108: system memory 122: primary host bridge

106: system bus 200: node

202: SMP processor 206: PHB

208: USB controller 210: PCI slot

300: CI / OC (300) 304: common I / O device

300: CI / OC 300 302: SP CI / OC 300 common I / O device 304.

400: CI / OC 406: Input

408: switch 404: common I / O device

504: keyboard 506: mouse

508: USB-ISA Conversion Logic 510: ISA Floppy Drive

512: serial port 514: parallel port

A preferred embodiment of the present invention provides a cabinet input / output controller (CI / OC) that allows multiple nodes in a multiprocessor system to share a common I / O device. In a preferred embodiment of the present invention, other run-time performance related I / Os, such as Ethernet adapters and their associated network connections, will continue to be maintained at each node.

Referring now to the drawings and in particular to FIG. 1, there is shown a block diagram of a data processing system in which a preferred embodiment of the present invention may be implemented. For example, data processing system 100 may be one of the computer server models available from International Business Machines Inc. of Armonk, NY. Data processing system 100 includes processors 101 and 102, which in an example embodiment are connected to level 2 caches 103 and 104, respectively, which level 2 caches 103 and 104 are in turn a system. The bus 106 is connected in turn.

System memory 108 and main host bridge (PHB) 122 are also connected to system bus 106. PHB 122 couples I / O bus 112 to system bus 106 and relays and / or converts data transactions from one bus to another. In an exemplary embodiment, the data processing system 100 includes a graphics adapter 118 connected to the I / O bus 112 for receiving user interface information for the display 120. Peripheral devices such as non-volatile storage 114, which may be a hard disk drive, and keyboard / pointing device 116, which may include a conventional mouse, trackball, etc. Is connected to the / O bus 112. PHB 122 is also connected to PCI slot 124 and universal serial bus controller (USB) 126 via I / O bus 112.

The exemplary embodiment shown in FIG. 1 is provided for the purpose of illustrating the invention only and those skilled in the art will recognize that many variations in the construction and functionality of FIG. 1 are possible. For example, data processing system 100 may also include a CD-ROM or digital video disk (DVD) drive, sound card, audio speakers, and other optional components. All such variations are intended to be included within the spirit and scope of the invention. The data processing system 100 and the CI / OC structure illustrated below are for illustrative purposes only and are not intended to impose structural limitations.

Referring now to FIG. 2, there is shown a node 200 that can be used as a basic construction of a large scale multiprocessor system. Node 200 includes an SMP processor 202 and an associated memory 204 (which may be shared with other processors). SMP processor 202 is connected to PHB 206, which is connected to USB controller 208 and PCI slot 210. Connected (or typically plugged in) to the PCI slot 210 is an I / O device 212, which is not shared with other nodes in this embodiment. The USB controller 208 preferably provides a node input connection to the CI / OC 216, which is controlled by the service processor (SP) 214. CI / OC 216 is connected to another node via node input 218 and allows node inputs 218 to share I / O device 220. Note that although this diagram shows a detailed description of only one example node, node 200 and other node 218 are each connected to a CI / OC by the same node input of CI / OC 216. Should be.

Large multiprocessor systems may be arranged in a number of small, independent partitions or may be arranged in NUMA or clusters. If a large multiprocessor system is arranged in NUMA or cluster, optional node interconnect hardware 222 may be used to obtain the desired interconnect configuration. CI / OC 216 in accordance with the preferred embodiment of the present invention is used to control nodes and to control and interconnect a single, common set of devices and these nodes. For example, a large multiprocessor system housed within a single rack would have included multiple computer nodes, but the entire rack would only require a single operator terminal, diskette drive, and the like. Preferably, the service processor (SP: 214) manages the CI / OC configuration. Each node includes a USB controller 208.

Referring now to FIG. 3, there is shown a high level diagram of a CI / OC in accordance with a preferred embodiment of the present invention. In FIG. 3, the CI / OC 300 is shown connected to its service processor 302. Exemplary node inputs 306 and 308 are shown, which allow connection of nodes as shown in FIG. 2. CI / OC 300 is also connected to common I / O device 304. There is only one CI / OC 300 and one SP 302 per large multiprocessor and all nodes share a common I / O device 304 connected downstream of the CI / OC 300. . The number of node inputs on CI / OC 300 may vary depending on the implementation.

Referring now to FIG. 4, the CI / OC 400 is shown in more detail as including a series of switches 408 that connect the node input 406 to the downstream I / O device 406. The SP 402 switches at most one node to a common I / O device connection 404. In a preferred embodiment of the present invention all connections are USB-compliant and the device is hot-pluggable so that the device can be attached or removed from the node while the node is operating. Activating switch 408 is equivalent to attaching a descending connection. Deactivating the switch is equivalent to removing the downstream connection. The SP 402 will connect the downstream device 404 to the node input 406 as needed.

Referring to Figure 5, various exemplary I / O devices are shown, as well as a common I / O system that includes basic I / O used to enable common I / O connections. In this figure, CI / OC 500 is connected to a hub which is a USB hub according to a preferred embodiment of the present invention. Hub 502 allows communication from CI / OC to be forwarded to any attached device. These devices may include a keyboard 504 or a mouse 506. If a raw USB device cannot be used to satisfy this condition, this I / O device may include commercially available USB-ISA conversion logic 508. By doing so, legacy devices such as ISA floppy drive 510, serial port 512, and parallel port 514 can be attached.

The Universal Serial Bus specification, available at http://www.usb.org, and incorporated herein by reference, provides the signals and signals on four wires of USB consisting of VBUS, GND, D +, and D-. It is detailed that power is transmitted. Signaling occurs on two wires D + and D-. According to a preferred embodiment of the present invention, the hub 502 is a powered hub that powers the attached device, so it does not require VBUS and GND from the node USB controller. In this embodiment, after the interconnection from node USB controller 208 to CI / OC as shown in FIG. 2, the interconnection to hub 502 as shown in FIG. 5 is omitted with VBUS and GNDUSB omitted. It consists of USB signals D + and D-.

Referring now to FIG. 6, a flow diagram for the preferred operation of a common I / O system is shown. When the node requests a common I / O device (step 600), the node sends a request to the service processor requesting an attach operation (step 610). If no node is currently using an I / O channel (step 620), the SP switches the CI / OC so that the node communicates with the downstream device (step 630). The node uses the I / O device (step 640) and, upon terminating it, instructs the SP to disconnect the CI / OC connection and detaches the I / O device (step 650). This node will then continue to operate normally (step 660).

If another node is using the I / O channel (step 620), the connection is denied (step 670). This node can retry to resume normal operation and establish as many connections as necessary. Due to the nature of multiprocessor systems, this kind of device conflict will be relatively exceptional and no more complex arbitration techniques are needed, but may be implemented by those skilled in the art if necessary.

While the invention has been shown and described with reference to particularly preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made without departing from the spirit and scope of the invention. For example, if the number of nodes is large, the CI / OC blocks may be connected in series so that each node may be connected to one CI / OC and communicate with an external device through a chain of CI / OCs. This and other variations are considered within the scope of the following claims.

The present invention provides a system and method for the operation of a multiprocessor computer system that improves resource allocation within a multiprocessor compute system and shares a common set of I / O devices among nodes of a multiprocessor computer system.

Claims (18)

In a computer system A plurality of system nodes, each having at least one system processor and associated memory, each of the system nodes being connected to communicate through respective device ports; A device controller connected to at least one of the device ports, At least one peripheral device connected to the device controller, The system is Sending a request from a system node to the device controller via a direct connection; Establishing an exclusive connection between the system node and the peripheral device by the controller; And operating the peripheral device by the system node. The method of claim 1, The system node operates as a symmetric multiprocessor system. The method of claim 1, Each of the system nodes sharing the peripheral device. The method of claim 1, The connection is a computer system in accordance with the Universal Serial Bus specification. The method of claim 1, And the system is further configured to perform a step of preventing a connection between a second system node and the peripheral device when the system node and the peripheral device are connected. The method of claim 1, The controller is connected to a plurality of computer systems, and the peripheral device is shared between the systems. The method of claim 1, The computer system is a rack-mounted system. The method of claim 1, And when multiple peripheral devices are connected, all of the peripheral devices are connected to the system during the setting step. In a multiprocessor computer system, A plurality of virtual computer systems each having at least one system processor and a memory connected to be written or read by the processor, A common input / output controller connected to the computer system, At least one peripheral device connected to the controller, The controller, when requested by the computer system, establishes an exclusive connection between one of the computer systems and the peripheral device. The method of claim 9, The virtual computer system is a partition within a symmetric multiprocessor system. The method of claim 9, The controller is connected to provide a connection to the peripheral device for the plurality of virtual computer systems. The method of claim 9, Each of the virtual computers is connected to the controller via a universal serial bus, and each peripheral device is connected to the controller via a universal serial bus hub. The method of claim 9, Each of the virtual computer systems comprises a universal serial bus controller. The method of claim 9, And when multiple peripheral devices are connected, all of the peripheral devices are connected to the virtual system when the peripheral devices are connected. In the device controller Communication controller, A plurality of node inputs, each operatively connected to said communication controller, A plurality of device connections, each operatively connected to said communication controller, The communication controller mediates node ownership of the device connection and provides an exclusive connection between both the given node input and the device connection when ownership of the device connection is granted to a given node input. The method of claim 15, And an integrated universal serial bus hub connected to the device connection. The method of claim 15, And the device connection is a universal serial bus connection. The method of claim 15, And an Industory standard Architecture device connection operatively connected to the communication controller.