WO2017029826A1 - リソース構成システム、リソース構成方法及びリソース構成プログラム - Google Patents
リソース構成システム、リソース構成方法及びリソース構成プログラム Download PDFInfo
- Publication number
- WO2017029826A1 WO2017029826A1 PCT/JP2016/057978 JP2016057978W WO2017029826A1 WO 2017029826 A1 WO2017029826 A1 WO 2017029826A1 JP 2016057978 W JP2016057978 W JP 2016057978W WO 2017029826 A1 WO2017029826 A1 WO 2017029826A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- resource
- calculation
- resource configuration
- provisioning
- processing
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/5044—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering hardware capabilities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
Definitions
- the present invention relates to a technology for configuring resources on the cloud.
- the present invention relates to a technology for selecting a resource that meets user requirements and a provisioning method for the resource in a cloud service that can use various resources and provisioning methods, and reconfiguring the resource configuration as necessary.
- cloud services that provision virtual resources of computer systems to users, such as IaaS (Infrastructure as a Service) type cloud services, have become widespread.
- a user can use virtual resources such as a virtual server, a virtual network, a virtual router, a virtual storage, and a virtual load balancer on demand by contracting with a service provider.
- a cloud service such as Amazon Web Services (registered trademark, http://aws.amazon.com/ec2).
- Users can run their own operating system (middleware such as OS (Operating System), DB (Data Base), Web, E-mail, etc.) on virtual resources, without preparing their own physical hardware devices. You can build a server.
- open source software for realizing IaaS is also popular.
- OpenStack registered trademark
- commercial cloud services such as Rackspace (registered trademark, Public Cloud Powered by OpenStack, http://www.rackspace.com/cloud/) exist.
- virtual servers provided to users in cloud services in the city when classified by the provisioning method, servers with many-core CPUs such as Xen and KVM (Kernel-based Virtual-Machine) Virtual servers virtualized with hypervisors are the mainstream.
- the hypervisor since the hypervisor has a drawback that the overhead of virtualization processing is large, there are container type virtual servers (containers) with little performance degradation and bare metal type physical servers (bare metal) that do not perform virtualization.
- FPGA Field Programmable Gate Array
- service providers have emerged that use a GPU (Graphical Processing Unit) that a general client PC has on a cloud server to provide a server with high image processing capabilities as a cloud server. For example, Amazon Web Service GPU instance server.
- virtual storage provided to users includes block storage and object storage when classified by the provisioning method.
- the mainstream format was to cut out virtual volumes from HDD-based dedicated storage (for example, EMC VNX) and provide them to users, but random IO (Input Output) performance is Service providers that offer SSDs and hybrid storage of SSDs and HDDs as virtual volumes are also emerging.
- EMC VNX dedicated storage
- random IO (Input Output) performance is Service providers that offer SSDs and hybrid storage of SSDs and HDDs as virtual volumes are also emerging.
- Yahoo registered trademark
- Nimble hybrid storage for cloud storage.
- distributed storage such as Ceph and Swift is used.
- Konoha of GMO cloud For example, Konoha of GMO cloud.
- Patent Document 1 it is conceivable to use the technique of Patent Document 1 or Non-Patent Document 1.
- Patent Document 1 is a technology that places a user's virtual resource on an appropriate physical device in a cloud service that provides the virtual resource.
- Non-Patent Document 1 is a technology for proposing an appropriate provisioning method according to the user's functional requirements and performance requirements in a cloud service provided with bare metal, containers, and hypervisors.
- a provisioning method is proposed, and virtual resources corresponding to the performance requirements and functional requirements of the user cannot be selected from virtual servers equipped with GPUs or FPGAs.
- Patent Literature 1 and Non-Patent Literature 1 are technologies for providing virtual resources to a user. When a user's usage mode for virtual resources changes, the configuration of the virtual resources is reconfigured accordingly. It is not possible.
- the present invention has been made in view of the above circumstances, and an object thereof is to provide a cloud service that realizes high processing performance specialized in specific processing, image processing, or parallel processing.
- a resource configuration system of the present invention includes a resource selection device that selects a resource on the cloud, and a resource reconfiguration device that configures the resource or reconfigures the resource configuration.
- the resource selection device requires receiving means for receiving a user's requirement for the resource, and the requirement requires specific processing, image processing, or parallel processing to be processed with a certain processing performance or more.
- a selection means for selecting a calculation resource and a provisioning method from a plurality of calculation resources including at least an FPGA or a GPU and a plurality of provisioning methods, respectively.
- the resource configuration method of the present invention is the resource selection method performed by a resource selection device that selects a resource on the cloud and a resource reconfiguration device that configures the resource or reconfigures the resource configuration.
- the apparatus receives at least a FPGA or a GPU based on receiving a user requirement for the resource and whether the requirement requires specific processing, image processing, or parallel processing to be performed with a processing performance of a certain level or more. Selecting a computing resource and a provisioning method from a plurality of computing resources and a plurality of provisioning methods, respectively.
- the gist of the resource configuration program of the present invention is to cause a computer to function as the resource configuration system according to claim 1.
- FIG. 1 is a diagram illustrating an overall configuration of a resource configuration system according to the first embodiment.
- This resource configuration system is arranged on a cloud system and selects a resource and a provisioning method that match a user's performance requirements and functional requirements, and an IaaS controller 3 that constructs a resource configuration based on the selection result , And is configured.
- the IaaS controller 3 is configured using, for example, an existing OpenStack.
- OpenStack is open source software for building a cloud computing infrastructure.
- the IaaS controller 3 can communicate with a plurality of physical devices arranged on the cloud, and manages the virtual resources on the cloud via the virtualization software such as Volume ⁇ ⁇ Manager and the physical resources without going through the virtualization software.
- an OpenStack API 31 that accepts a setting request from the resource selection device 1, a resource configuration building unit 32 (Heat, Nova, Cinder, etc.) that builds the resource configuration of the user based on the setting request,
- OpenStack DB 33 for storing various resource information existing in the server and resource configuration information for each user.
- This resource selection device 1 operates between the user terminal 5 and the IaaS controller 3.
- a request reception unit 11 that receives a user's performance requirements and functional requirements input at the user terminal 5, a resource selection unit 12 that selects a resource and a provisioning method based on the received request, and the selection result
- a selection result notifying unit 13 for notifying the user terminal 5, a setting requesting unit 14 for requesting resource setting to the IaaS controller 3 based on a user's approval instruction for the selection result, and information used when selecting a resource and a provisioning method are stored.
- the information used at the time of selection is, for example, information on all resources and provisioning methods that can be set and used by the IaaS controller 3, requests from the user terminal 5, selection results of resources and provisioning methods based on the requests, and the like.
- the resource selection device 1 can be realized by a computer having a calculation function such as a CPU and a storage function such as a memory. It is also possible to create a program for causing the computer to function and a storage medium for the program. Further, the resource selection device 1 may function as a device physically different from the IaaS controller 3 or may be a function of the IaaS controller 3.
- a resource refers to either or both of a virtual resource and a physical resource.
- OpenStack when used as the IaaS controller 3, generally a virtual resource using virtualization software is conceivable, but a physical resource may be selected as a selection target resource without using the virtualization software.
- the provisioning method will be described.
- the resource is a calculation resource, for example, bare metal provisioning, container provisioning, and hypervisor provisioning are used.
- the resource is a storage resource, for example, block storage provisioning or object storage provisioning is used.
- Bare metal is a physical server that does not perform virtualization. Same as conventional Dedicated Hosting. Therefore, bare metal provisioning means securing a physical server that does not perform virtualization. For example, in OpenStack, a component called Ironic performs bare metal provisioning. Since bare metal is a dedicated server, it has a high degree of freedom and high performance, but it takes a long time to provision and start up, and live migration cannot be performed.
- a container is a virtual server using OS virtualization technology. Used in VPS (Virtual Private Server) such as OpenVZ. Therefore, container provisioning refers to securing a virtual server using OS virtualization technology.
- Container provisioning virtualizes physical servers by isolating computer resources in units called containers. Container provisioning differs from the hypervisor described later in that it shares the OS kernel. Docker using LXC (LinuxCContainer) appeared in 2013, and its usage is increasing due to its ease of use. The container does not have the degree of freedom of the kernel, but since the securing and generation processing is just a process startup, the startup time is short and the performance degradation is small. OpenVZ can be live migrated, but Docker and LXC are not.
- Hypervisor is a technology that virtualizes a physical server using hardware virtualization technology. Therefore, hypervisor provisioning means securing a virtual server using hardware virtualization technology. Since hypervisor provisioning operates on hardware emulated with hardware virtualization technology, the OS can be freely customized. Examples of main hypervisors include Xen, KVM, and VMware ESX. This virtual server has a high degree of freedom of the OS and can be live migrated. On the other hand, since this virtual server has a large emulation overhead, it has a disadvantage that it is inferior to bare metal or containers in terms of performance and startup time.
- calculation resources will be described.
- a CPU, GPU, or FPGA is used.
- the CPU is a general computing resource installed in many servers and client PCs, and is designed with low latency in mind.
- the CPU has a large cache, advanced control, and powerful arithmetic functions.
- cloud servers often use many-core CPUs with more than 10 CPU cores.
- GPU is a calculation unit for image processing and is designed with high throughput in mind.
- the GPU has a small cache, simple control, energy efficient computing functions, and compute units are arranged to perform a large amount of parallel computation for high throughput.
- Many GPUs are installed in client PCs, but they are also installed in servers that run applications for image processing.
- FPGA is an integrated circuit whose configuration can be set or changed by a purchaser or designer after manufacture, and is one of programmable logic devices. FPGAs have advantages such as the ability to update functions after shipment, partial reconfiguration in terms of design, and lower engineering costs than ASIC design. FPGA is not suitable for general-purpose processing, but it can achieve higher processing performance than CPU and GPU by configuring logic for specific calculation. If it is the same processing suitable for FPGA, it has processing performance several times that of GPU and 10-100 times that of CPU. FPGAs were mainly used for scientific and academic purposes, but cloud servers are also being used to speed up search processing and NoSQL Engine processing. For example, it is used by Microsoft's Bing search and IBM's NoSQL engine.
- HDD high-disk
- SSD solid state drive
- hybrid storage high-speed storage
- HDD Hard Disk Drive
- striping is done by parallelizing HDDs to improve IO (Input Output) performance.
- IO Input Output
- a typical 150001rpm (rotation per minute) HDD is about 150-200 IOPS with random write, but it becomes about 3000-4000 IOPS by arranging 20 HDDs.
- power and space costs are required, and when an HDD is added or an HDD failure is replaced, the service must be stopped due to the storage chassis being stopped.
- SSD Solid State Drive
- IOPS Online Transaction Processing
- Hybrid storage is a storage resource that takes advantage of the features of each HDD and SSD.
- Nimble's Storage employs a method called CAS ((Cache Accelerated Sequential Layout).
- CASL when writing, first write to SSD, compress to a format that can be accessed sequentially, write to HDD, and at the same time, part of it is saved in cache.
- search from SSD and cache When reading, search from SSD and cache, finally search HDD and return response.
- CASL achieves random IOPS that is at least as good as regular SSDs.
- Research on hybrid storage has been active since the end of the 2000s, and there are many other research and products besides Nimble. For example, see “Extending SSD Lifetimes Disk-Based Write Caches” (G. Soundararajan, 3 others, FAST 2010: 8th USENIX Conference on File Storage Technologies, Feb. 2010.).
- Distributed storage is a storage resource that uses an IA (Intel Architecture) server.
- IA Intelligent Architecture
- Large-capacity and scalable storage systems have been built by placing inexpensive IA servers in parallel.
- SPOF Single Point Of Of Failure
- SSDs can be used as drives, but SATA (Serial Advanced Technology Attachment) HDDs are usually used with emphasis on cost.
- the software that configures the distributed storage can place multiple copies of one data on multiple nodes, and can automatically perform rearrangement at the time of addition and degeneration at the time of failure.
- Examples of distributed storage used as a file system include GlusterFS, Ceph, GFS, and HDFS.
- there is OpenStackftSwift as an object storage, for example, there is OpenStackftSwift.
- the resources that can be selected by the resource selection device 1 and the provisioning method have been described.
- the characteristics and performance of such resources and provisioning methods are compared and examined for each resource and each provisioning method.
- a certain requirement criterion is defined hierarchically for the performance requirement and functional requirement required by the user, and the resource and the provisioning method are specified according to the determination result in each criterion, and the user requirement Select matching resources and provisioning methods. This makes it possible to place resources and provisioning methods according to user requirements in the right place and provide processing performance resources and provisioning methods that meet user requirements.
- FIG. 2 is a diagram illustrating a processing flow for selecting a resource and a provisioning method.
- the cloud provider proposes a resource and a provisioning method.
- the user has little know-how regarding performance design and device selection.
- the request reception unit 11 receives performance requirements and functional requirements specified by the user on the user terminal 5 (step S101).
- PaaS Platinum ⁇ as a Service
- PaaS is a platform that manages the execution of applications on the cloud. It is also possible to deploy an application via the PaaS. For example, Cloud Foundry is used.
- the functional requirements are information on whether the OS is Linux or other than Linux, whether kernel customization is required, what kind of application runs, and the like.
- the performance requirement is information related to throughput, latency, and the like.
- the resource selection unit 12 selects a resource and a provisioning method that match the performance requirement and the functional requirement specified in Step S101 from the plurality of resources and the plurality of provisioning methods (Step S102). Specifically, in consideration of the features and performance of the resource and the provisioning method that can be selected by the resource selection device 1, necessary condition determination criteria based on the features and performance are hierarchically defined for user requirements. Then, the specified performance requirement and functional requirement are determined by each necessary condition determination criterion, and the resource and provisioning method corresponding to the determination result are selected.
- provisioning methods include bare metal, containers, or hypervisors, and server equipment includes normal CPU-based servers, GPU-enhanced servers, and FPGAs optimized for specific computing processes. Select one of the selected servers.
- the provisioning method is either block storage or object storage
- the storage device is either HDD-based storage, SSD-based storage, HDD-SSD hybrid storage, or distributed storage such as Ceph. Select. A specific selection procedure example of resources and provisioning methods will be described later.
- the selection result notifying unit 13 transmits the resource and the provisioning method selected in step S102 to the user terminal 5 (step S103). Receiving it, the user terminal 5 displays the received resource and the provisioning method on the screen as the proposed resource and the proposed provisioning method. The user of the user terminal 5 confirms the content of the proposal, and returns an approval instruction to the resource selection device 1 if there is no problem.
- the setting request unit 14 specifies the resource approved by the user and the provisioning method, and requests the IaaS controller 3 for resource setting and generation (step S104). Thereafter, the IaaS controller 3 secures resources based on the request from the resource selection device 1.
- step S102 a specific selection procedure of resources and provisioning methods performed in step S102 will be described.
- a procedure for selecting a calculation resource will be described with reference to FIG.
- step S102a it is determined whether or not the user's functional requirements and performance requirements require that a specific calculation process be processed with a certain processing performance.
- step S102a If the determination result in step S102a is Yes, a server including an FPGA that accelerates the specific calculation process is selected, and bare metal is selected as a provisioning method for the server (step S102b).
- Examples of speeding up in FPGA include speeding up memcached and speeding up Microsoft's Bing search. Provision of bare metal is necessary because the benefits of FPGA cannot be fully utilized when virtualized. The method for speeding up memcached is described in “Thin Servers with Smart Pipes: Designing SoC Accelerators for Memcached”, http://bit.ly/1BBNBEI, ISCA, 2013 ”.
- step S102a determines whether the user's functional requirements and performance requirements require that image processing (or parallel processing) be processed with a certain level of processing performance or more. (Step S102c).
- step S102c determines whether a server with enhanced GPU is selected, and a bare metal or container is selected as the provisioning method for the server (steps S102e and S102f). Which provisioning method to select is determined based on the level of processing performance required for the computing resource or the necessity of customization for the OS (step S102d). If the determination result in step S102d is Yes, bare metal is selected, and if it is No, a container is selected. Note that image processing includes applications that perform video editing, processing, analysis, and the like. Server GPUs are often not as strong as client PCs, but servers with powerful GPUs are emerging. When controlling the GPU, the hardware is abstracted and cannot be controlled in the virtual machine, so provisioning is performed with bare metal or a container.
- containers are often limited to Linux, such as LXC and Docker, provisioning with bare metal is required for other than Linux.
- container technology that Windows is compatible with the OS has recently appeared, and by using it, container provisioning is possible even when the OS is Windows.
- step S102c determines whether a normal CPU-based server is selected, and a bare metal, hypervisor, or container is selected as a provisioning method for the server (steps S102i, S102j, S102k).
- the provisioning method to be selected is determined based on the size of processing performance required for the computing resource (for example, high throughput, low delay) or the necessity of customization for the OS (steps S102g, S102h). ). If the determination result in step S102g is Yes, bare metal is selected. On the other hand, if the determination result is No and the determination result in Step S102h is Yes, a hypervisor is selected. If both of the determination results in step S102g and step S102h are No, a container is selected.
- step S102c is performed before step S102a.
- a storage resource necessary condition determination criteria based on characteristics and performance of the storage resource and the provisioning method are hierarchically determined, and a calculation resource and a provisioning method that match the specified performance requirement and functional requirement are selected.
- block storage or object storage is selected as a provisioning method based on the functional requirements specified by the user.
- Block storage manages data in units of blocks.
- Object storage is managed in units of objects. Since object storage is generally inexpensive, if the object storage is specified by the user as a functional requirement, a distributed storage such as OpenStack Swift is selected.
- block storage is designated as a functional requirement, the storage is selected according to the characteristics of the application program to be operated.
- HDD-SSD hybrid storage For applications that require high IO, such as OLAP, select SSD or HDD-SSD hybrid storage according to performance requirements. On the other hand, for applications that do not require high IO such as file servers and archives, select HDD or distributed storage according to performance requirements. Generally, the price is higher in the order of SSD> HDD-SSD hybrid> HDD> distributed storage, and an inexpensive storage that satisfies the performance requirements is selected.
- whether the resource selection device 1 requires that the user's performance requirement and functional requirement process specific calculation processing, image processing, or parallel processing with a certain level of processing performance or more Based on a plurality of calculation resources including at least FPGA and GPU, a plurality of storage resources, and a plurality of provisioning methods, a predetermined calculation resource, a predetermined storage resource, and a predetermined provisioning method are selected respectively.
- a cloud service that realizes high processing performance specialized for parallel processing, and further, it is possible to reduce operation and labor such as server configuration design by the user.
- the resource selection device 1 selects a storage resource according to the characteristics of the application program to be operated. Therefore, such as OLAP that requires high IO while HDD, SSD, etc. are mixed. Storage with high IO such as SSD can be presented to users who want to perform processing.
- the resource selection device 1 includes an FPGA optimized for the calculation when it is desired to perform a specific calculation process at high speed while a CPU, a GPU, an FPGA, and the like are mixed. If you want to provision hardware with bare metal and want to perform image processing at high speed, provision hardware with GPU enhancement with bare metal or container, otherwise use a normal CPU server with bare metal, container, or hypervisor. Since provisioning is performed, the user can realize high processing performance on the computing resource.
- FIG. 4 is a diagram showing an overall configuration of the resource configuration system according to the second embodiment.
- This resource configuration system includes a resource reconfiguration device 7 that configures and reconfigures resources and provisioning methods, and an IaaS controller 3 that constructs or reconstructs a resource configuration based on the configuration or reconfiguration result. .
- the resource reconfiguration device 7 includes a request reception unit 71 that receives a user performance requirement and a functional requirement input from the user terminal 5, and a resource reconfiguration unit 72 that configures or reconfigures a resource and a provisioning method based on the received request.
- a reconfiguration result notifying unit 73 for notifying the user terminal 5 of the configuration or reconfiguration result, and a setting requesting unit 74 for requesting resource setting to the IaaS controller 3 based on a user approval instruction for the configuration or reconfiguration result;
- a resource usage frequency collection unit 75 that collects user usage frequency with respect to the configured or reconfigured resource, and a data storage unit 76 that stores information used when configuring and reconfiguring the resource and the provisioning method. Is done.
- the resource reconfiguration device 7 can be realized by a computer having a calculation function such as a CPU and a storage function such as a memory. It is also possible to create a program for causing the computer to function and a storage medium for the program. Furthermore, the resource reconfiguration device 7 may function as a device physically different from the resource selection device 1 or the IaaS controller 3, or may be a function of the resource selection device 1 or the IaaS controller 3.
- This operation example will be described using two operation patterns.
- the first is the operation that constitutes the calculation logic of the FPGA.
- the second is an operation of reconfiguring the calculation logic of the FPGA during the user's resource operation.
- FIG. 5 is a diagram showing a configuration processing flow of calculation logic for the FPGA server.
- the request receiving unit 71 receives performance requirements and functional requirements specified by the user on the user terminal 5 (step S201).
- the functional requirements are information regarding whether calculation logic that needs to be speeded up is necessary, what kind of application operates, and the like.
- the performance requirement is information related to throughput, latency, and the like.
- the resource reconfiguration unit 72 configures the calculation logic of the FPGA server based on the performance requirement and functional requirement specified in step S201, and further selects a provisioning method (step S202). Specifically, an FPGA server that is not being used is selected, and calculation logic is set according to the specified performance requirements and functional requirements. If a calculation logic that requires high speed is not specified, it may be configured according to the type of application.
- the configuration process itself can be realized using a known technique. For example, the acceleration by FPGA in NoSQL is described in Data Engine for NoSQL-IBM Power Systems Edition White Paper. What is necessary is just to define in advance inside the resource reconfiguration apparatus 7 to set to the structure based on this description content.
- the time required for the configuration or reconfiguration of the FPGA is generally about several tens of ms, and there is no problem even if the configuration or reconfiguration is performed according to the user requirements.
- a specific configuration procedure example of the calculation logic for the FPGA server will be described later.
- bare metal is selected as in step S102b of FIG.
- the reconfiguration result notifying unit 73 transmits the calculation logic to the FPGA configured in step S202 and the selected provisioning method to the user terminal 5 (step S203).
- the user terminal 5 displays the received calculation logic and provisioning method on the screen as the proposed resource and the proposed provisioning method.
- the user of the user terminal 5 confirms the content of the proposal, and returns an approval instruction to the resource selection device 1 if there is no problem.
- the calculation logic set in the FPGA may be designated by the user and returned.
- the setting request unit 74 specifies the resource approved by the user and the provisioning method, and requests the IaaS controller 3 to set and generate the resource (step S204). Thereafter, the IaaS controller 3 secures resources based on the request from the resource selection device 1. As a result, the calculation logic and provisioning method configured and selected in step S202 are used in the unused FPGA server.
- FIG. 6 is a diagram showing a configuration processing flow of calculation logic for the FPGA server.
- step S202a it is determined whether or not the user's functional requirements and performance requirements specify a specific calculation logic.
- step S202a If the determination result in step S202a is Yes, the FPGA server is configured using the specified specific calculation logic (step S202b). On the other hand, when the determination result of step S202a is No, it is determined whether or not the calculation logic corresponding to the application type specified by the user can be used on the resource configuration system (step S202c).
- step S202c determines whether the FPGA server is configured using the calculation logic corresponding to the designated application type.
- step S202e determines whether the calculation logic corresponding to the designated application type.
- the specified configuration in the FPGA for example, it may be reflected using a development tool provided by Altera or Xilinx developing the FPGA.
- FIG. 7 is a diagram showing a reconfiguration processing flow of calculation logic for the FPGA server. This second operation is performed, for example, after the first operation.
- the resource usage frequency collection unit 75 periodically collects usage frequencies of various calculation processes used by the user in each FPGA server (step S301). For example, the number of graph analyzes is collected for each FPGA server, such as how many times the graph analysis is performed per hour, and how many times NoCR CRUD processing is performed per hour.
- the resource reconfiguration unit 72 detects the specific calculation processing of each FPGA server based on the type of calculation logic set in the FPGA server, and the type of calculation processing that is detected in each FPGA server.
- the number of times of various calculation processes in each FPGA server is periodically collated, and it is determined whether it is better to reconfigure the FPGA server based on the collation result (step S302). For example, on an FPGA server that is good at graph analysis, if graph analysis is performed 10 times per hour and NoSQL CRUD processing is performed 10,000 times per hour, it is reconfigured to an FPGA server that is good at NoSQL processing. It is better to do.
- step S302 for example, when the number of calculations other than the good calculation exceeds a certain threshold, it is determined that the FPGA server is properly reconfigured.
- An FPGA server that is good at graph analysis is an FPGA server on which calculation logic for graph analysis is set.
- the reconfiguration result notifying unit 73 transmits a notification that the calculation logic should be changed to the user terminal 5 (step S303). For example, in the case of the above example, it is suggested to the user that it is more appropriate to change the calculation logic for graph analysis set in the corresponding FPGA server to the calculation logic for NoSQL. Receiving it, the user terminal 5 displays the received calculation logic change proposal on the screen. The user of the user terminal 5 confirms the proposal and returns an approval instruction to the resource selection device 1 if there is no problem. At this time, if there is a problem, the calculation logic to be set in the FPGA may be specified by the user and returned, or an instruction not to reconfigure may be returned.
- the resource reconfiguration unit 72 selects the FPGA server used by the user, and based on the approval result by the user, the calculation logic of the FPGA server is increased so that the calculation logic with the increased number of calculations is accelerated. Change (step S304).
- the calculation logic for graph analysis is changed to the calculation logic for NoSQL whose usage frequency has increased.
- the time required for FPGA reconfiguration is generally about several tens of ms, and there is no problem even if the FPGA server is reconfigured during operation of the FPGA server.
- This second operation pattern may be applied to the FPGA server that is provisioned by the resource selection device 1 used in the first embodiment and for which a specific calculation process is optimized in advance.
- the resource reconfiguration device 7 changes the setting to the FPGA server optimized for the specific calculation logic and performs the bare metal provisioning according to the user's requirement, so that the user has high processing performance. realizable.
- the resource reconfiguring device 7 collects the usage frequencies of various calculation processes used by the user on the FPGA server.
- specific calculation processing increases due to changes in user usage
- the configuration of the FPGA server is changed to match the increased calculation processing, so the user can achieve higher processing performance on the FPGA server, and more The need to prepare a plurality of FPGA servers with calculation logic in advance can be reduced.
- Non-Patent Document 1 includes a mechanism for automatically verifying the performance of the server configuration proposed by the service provider side, and the user can easily perform the performance verification by using them.
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Stored Programmes (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
図1は、第1の実施の形態に係るリソース構成システムの全体構成を示す図である。このリソース構成システムは、クラウドシステム上に配置され、ユーザの性能要件及び機能要件に合致するリソース及びプロビジョニング方法を選択するリソース選択装置1と、その選択結果に基づきリソース構成を構築するIaaSコントローラ3と、を備えて構成される。
第1の実施の形態では、リソース及びプロビジョニング方法の選択方法について説明した。一方、第2の実施の形態では、ユーザの要求に応じてリソースを選択及び構成し、さらにユーザによるリソースの利用方法の変化に応じて該リソース構成を再構成する方法について説明する。
11…要求受付部
12…リソース選択部
13…選択結果通知部
14…設定要求部
15…データ記憶部
3…IaaSコントローラ
31…OpenStack API
32…リソース構成構築部
33…OpenStack DB
5…ユーザ端末
7…リソース再構成装置
71…要求受付部
72…リソース再構成部
73…再構成結果通知部
74…設定要求部
75…リソース利用頻度収集部
76…データ記憶部
S101~S104、S102a~S102k、S201~S204、S202a~S202e、S301~S304…ステップ
Claims (9)
- クラウド上のリソースを選択するリソース選択装置と、リソースを構成し又はそのリソース構成を再構成するリソース再構成装置と、を備えたリソース構成システムにおいて、
前記リソース選択装置は、
前記リソースに対するユーザの要件を受信する受信手段と、
前記要件が特定処理、画像処理、又は並列処理を一定以上の処理性能で処理することを要求しているかに基づき、少なくともFPGAかGPUを含む複数の計算リソース及び複数のプロビジョニング方法から計算リソース及びプロビジョニング方法をそれぞれ選択する選択手段と、
を備えることを特徴とするリソース構成システム。 - 前記選択手段は、
前記要件が前記特定処理を一定以上の処理性能で処理することを要求している場合、前記計算リソースとしてFPGAを選択し、当該計算リソースのプロビジョニング方法としてベアメタルを選択することを特徴とする請求項1に記載のリソース構成システム。 - 前記選択手段は、
前記要件が前記画像処理又は前記並列処理を一定以上の処理性能で処理することを要求している場合、前記計算リソースとしてGPUを選択し、当該計算リソースのプロビジョニング方法として、計算リソースに対して要求されている処理性能の大きさ又はOSに対するカスタマイズの要否に基づきベアメタル又はコンテナを選択することを特徴とする請求項1に記載のリソース構成システム。 - 前記選択手段は、
前記要件に基づきブロック単位又はオブジェクト単位のプロビジョニング方法を選択し、動作させるアプリケーションプログラムの特性に応じて記憶リソースを選択することを特徴とする請求項1に記載のリソース構成システム。 - 前記リソース再構成装置は、
ユーザの要件に応じた計算ロジックをFPGAの計算リソースに設定し、当該計算ロジックを当該計算リソースで利用させる構成手段、
を備えることを特徴とする請求項1に記載のリソース構成システム。 - 前記リソース再構成装置は、
リソース構成後の計算リソースで前記ユーザが利用している各種計算処理の利用頻度をそれぞれ収集する収集手段と、
前記利用頻度が増大した特定の計算処理に合わせて前記リソース構成後の計算リソースの構成を変更する再構成手段と、
を備えることを特徴とする請求項1に記載のリソース構成システム。 - クラウド上のリソースを選択するリソース選択装置と、リソースを構成し又はそのリソース構成を再構成するリソース再構成装置と、で行うリソース構成方法において、
前記リソース選択装置は、
前記リソースに対するユーザの要件を受信するステップと、
前記要件が特定処理、画像処理、又は並列処理を一定以上の処理性能で処理することを要求しているかに基づき、少なくともFPGAかGPUを含む複数の計算リソース及び複数のプロビジョニング方法から計算リソース及びプロビジョニング方法をそれぞれ選択するステップと、
を備えることを特徴とするリソース構成方法。 - 前記リソース再構成装置は、
リソース構成後の計算リソースで前記ユーザが利用している各種計算処理の利用頻度をそれぞれ収集するステップと、
前記利用頻度が増大した特定の計算処理に合わせて前記リソース構成後の計算リソースの構成を変更するステップと、
を備えることを特徴とする請求項7に記載のリソース構成方法。 - 請求項1に記載のリソース構成システムとしてコンピュータを機能させることを特徴とするリソース構成プログラム。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/742,765 US10628225B2 (en) | 2015-08-18 | 2016-03-14 | Resource configuration system, resource configuration method and resource configuration program for selecting a computational resource and selecting a provisioning method |
CN201680048125.9A CN107924340A (zh) | 2015-08-18 | 2016-03-14 | 资源配置系统、资源配置方法以及资源配置程序 |
JP2017535249A JP6438144B2 (ja) | 2015-08-18 | 2016-03-14 | リソース構成システム、リソース構成方法及びリソース構成プログラム |
EP16836818.1A EP3340051A4 (en) | 2015-08-18 | 2016-03-14 | Resource Configuration System, Resource Configuration Method, and Resource Configuration Program |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2015160784 | 2015-08-18 | ||
JP2015-160784 | 2015-08-18 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017029826A1 true WO2017029826A1 (ja) | 2017-02-23 |
Family
ID=58051546
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2016/057978 WO2017029826A1 (ja) | 2015-08-18 | 2016-03-14 | リソース構成システム、リソース構成方法及びリソース構成プログラム |
Country Status (5)
Country | Link |
---|---|
US (1) | US10628225B2 (ja) |
EP (1) | EP3340051A4 (ja) |
JP (1) | JP6438144B2 (ja) |
CN (1) | CN107924340A (ja) |
WO (1) | WO2017029826A1 (ja) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108958910A (zh) * | 2018-05-21 | 2018-12-07 | 福建省数字福建云计算运营有限公司 | 一种基于异构环境下的任务调度方法及终端 |
JP2019121185A (ja) * | 2018-01-05 | 2019-07-22 | コニカミノルタ株式会社 | Gpu割当プログラム、gpu割当方法、コンピュータ読取可能な記録媒体、および、gpu割当装置 |
WO2021151352A1 (zh) * | 2020-10-23 | 2021-08-05 | 平安科技(深圳)有限公司 | 裸金属服务交付方法、装置、计算机设备和存储介质 |
JP2021521533A (ja) * | 2018-04-18 | 2021-08-26 | ザイリンクス インコーポレイテッドXilinx Incorporated | 異種システムオンチップのためのソフトウェア定義マルチドメイン作成分離 |
WO2023105670A1 (ja) * | 2021-12-08 | 2023-06-15 | 日本電信電話株式会社 | リソース管理装置及びプログラム |
WO2023105671A1 (ja) * | 2021-12-08 | 2023-06-15 | 日本電信電話株式会社 | 計算機及びプログラム |
JP7571414B2 (ja) | 2020-08-06 | 2024-10-23 | 日本電気株式会社 | コンテナ実行システム、端末装置、サーバ装置、コンテナ実行環境決定方法及びプログラム |
Families Citing this family (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10484242B2 (en) * | 2017-08-23 | 2019-11-19 | Bank Of America Corporation | Dynamic cloud stack configuration |
US11140032B2 (en) * | 2018-05-04 | 2021-10-05 | Oracle International Corporation | System and method for building idempotent configuration management modules for a cloud Infrastructure Service |
CN108897601A (zh) * | 2018-06-29 | 2018-11-27 | 郑州云海信息技术有限公司 | 一种基于虚拟化的fpga使用方法、系统及相关装置 |
US11416663B2 (en) * | 2018-12-13 | 2022-08-16 | Hamilton Sundstrand Corporation | Optimized allocation of functions in hybrid motor controller implementations |
US11120188B2 (en) * | 2019-01-28 | 2021-09-14 | King Fahd University Of Petroleum And Minerals | FPGA virtualization |
CN111309440B (zh) * | 2020-02-16 | 2022-07-08 | 苏州浪潮智能科技有限公司 | 一种多类型gpu的管理调度的方法和设备 |
CN111858241B (zh) * | 2020-07-10 | 2022-06-10 | 浪潮电子信息产业股份有限公司 | 一种fpga云平台的监控方法、装置、设备及存储介质 |
US11809912B2 (en) | 2020-12-09 | 2023-11-07 | Dell Products L.P. | System and method for allocating resources to perform workloads |
US11928515B2 (en) | 2020-12-09 | 2024-03-12 | Dell Products L.P. | System and method for managing resource allocations in composed systems |
US11934875B2 (en) | 2020-12-09 | 2024-03-19 | Dell Products L.P. | Method and system for maintaining composed systems |
US11809911B2 (en) | 2020-12-09 | 2023-11-07 | Dell Products L.P. | Resuming workload execution in composed information handling system |
US11704159B2 (en) | 2020-12-09 | 2023-07-18 | Dell Products L.P. | System and method for unified infrastructure architecture |
US11853782B2 (en) | 2020-12-09 | 2023-12-26 | Dell Products L.P. | Method and system for composing systems using resource sets |
US11604595B2 (en) | 2020-12-09 | 2023-03-14 | Dell Products L.P. | Data mirroring and data migration between storage volumes using system control processors |
US11675625B2 (en) | 2020-12-09 | 2023-06-13 | Dell Products L.P. | Thin provisioning of resources using SCPS and a bidding system |
US11675665B2 (en) | 2020-12-09 | 2023-06-13 | Dell Products L.P. | System and method for backup generation using composed systems |
US11693703B2 (en) | 2020-12-09 | 2023-07-04 | Dell Products L.P. | Monitoring resource utilization via intercepting bare metal communications between resources |
US11435814B2 (en) | 2020-12-09 | 2022-09-06 | Dell Produts L.P. | System and method for identifying resources of a composed system |
US11698821B2 (en) | 2020-12-09 | 2023-07-11 | Dell Products L.P. | Composable information handling systems in an open network using access control managers |
US11755433B2 (en) | 2020-12-22 | 2023-09-12 | EMC IP Holding Company LLC | Method and system for health rank based virtual machine restoration using a conformal framework |
US11687280B2 (en) | 2021-01-28 | 2023-06-27 | Dell Products L.P. | Method and system for efficient servicing of storage access requests |
US11768612B2 (en) | 2021-01-28 | 2023-09-26 | Dell Products L.P. | System and method for distributed deduplication in a composed system |
US11675916B2 (en) | 2021-01-28 | 2023-06-13 | Dell Products L.P. | Method and system for limiting data accessibility in composed systems |
US11797341B2 (en) | 2021-01-28 | 2023-10-24 | Dell Products L.P. | System and method for performing remediation action during operation analysis |
US12013768B2 (en) | 2021-07-22 | 2024-06-18 | Dell Products L.P. | Method and system for automated healing of hardware resources in a composed information handling system |
US12026557B2 (en) | 2021-07-22 | 2024-07-02 | Dell Products L.P. | Method and system for a utilizing a proxy service to generate a composed information handling system |
US11947697B2 (en) | 2021-07-22 | 2024-04-02 | Dell Products L.P. | Method and system to place resources in a known state to be used in a composed information handling system |
US12008412B2 (en) | 2021-07-28 | 2024-06-11 | Dell Products | Resource selection for complex solutions |
US11928506B2 (en) | 2021-07-28 | 2024-03-12 | Dell Products L.P. | Managing composition service entities with complex networks |
CN115499305B (zh) * | 2022-07-29 | 2024-04-26 | 天翼云科技有限公司 | 一种分布式集群存储设备的部署方法、装置及电子设备 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2011248503A (ja) * | 2010-05-25 | 2011-12-08 | Nippon Telegr & Teleph Corp <Ntt> | 情報サービスシステム |
WO2014073024A1 (en) * | 2012-11-09 | 2014-05-15 | Hitachi, Ltd. | Management computer, computer system, and instance management method |
JP2015132887A (ja) * | 2014-01-09 | 2015-07-23 | 富士通株式会社 | 要求分散プログラム、要求分散方法および情報処理装置 |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6915518B1 (en) * | 2000-07-24 | 2005-07-05 | Xilinx, Inc. | System and method for runtime reallocation of PLD resources |
US20030069974A1 (en) * | 2001-10-08 | 2003-04-10 | Tommy Lu | Method and apparatus for load balancing web servers and virtual web servers |
US7437721B2 (en) * | 2004-09-29 | 2008-10-14 | Microsoft Corporation | Isolating software deployment over a network from external malicious intrusion |
US8799431B2 (en) * | 2005-08-15 | 2014-08-05 | Toutvirtual Inc. | Virtual systems management |
US9235442B2 (en) * | 2010-10-05 | 2016-01-12 | Accenture Global Services Limited | System and method for cloud enterprise services |
CN103312727A (zh) * | 2012-03-06 | 2013-09-18 | 创业软件股份有限公司 | 一种区域医疗协同服务领域云计算部署方法 |
US8892945B2 (en) * | 2012-04-02 | 2014-11-18 | International Business Machines Corporation | Efficient application management in a cloud with failures |
US8775576B2 (en) * | 2012-04-17 | 2014-07-08 | Nimbix, Inc. | Reconfigurable cloud computing |
DE102012221355A1 (de) * | 2012-11-22 | 2014-05-22 | Siemens Aktiengesellschaft | Verfahren zum Bereitstellen von Ressourcen in einer Cloud und Vorrichtung |
US8677359B1 (en) * | 2013-03-14 | 2014-03-18 | Joyent, Inc. | Compute-centric object stores and methods of use |
US20140280964A1 (en) * | 2013-03-15 | 2014-09-18 | Gravitant, Inc. | Systems, methods and computer readable mediums for implementing cloud service brokerage platform functionalities |
CN103297504B (zh) * | 2013-05-09 | 2016-05-11 | 浙江大学 | 一种云数据中心中物理裸机快速部署操作系统的方法 |
JP5988505B2 (ja) | 2013-11-26 | 2016-09-07 | 日本電信電話株式会社 | 仮想リソース管理装置、選択方法及び選択プログラム |
-
2016
- 2016-03-14 JP JP2017535249A patent/JP6438144B2/ja active Active
- 2016-03-14 WO PCT/JP2016/057978 patent/WO2017029826A1/ja active Application Filing
- 2016-03-14 EP EP16836818.1A patent/EP3340051A4/en not_active Ceased
- 2016-03-14 US US15/742,765 patent/US10628225B2/en active Active
- 2016-03-14 CN CN201680048125.9A patent/CN107924340A/zh active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2011248503A (ja) * | 2010-05-25 | 2011-12-08 | Nippon Telegr & Teleph Corp <Ntt> | 情報サービスシステム |
WO2014073024A1 (en) * | 2012-11-09 | 2014-05-15 | Hitachi, Ltd. | Management computer, computer system, and instance management method |
JP2015132887A (ja) * | 2014-01-09 | 2015-07-23 | 富士通株式会社 | 要求分散プログラム、要求分散方法および情報処理装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP3340051A4 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019121185A (ja) * | 2018-01-05 | 2019-07-22 | コニカミノルタ株式会社 | Gpu割当プログラム、gpu割当方法、コンピュータ読取可能な記録媒体、および、gpu割当装置 |
JP6992515B2 (ja) | 2018-01-05 | 2022-01-13 | コニカミノルタ株式会社 | Gpu割当プログラム、gpu割当方法、コンピュータ読取可能な記録媒体、および、gpu割当装置 |
JP2021521533A (ja) * | 2018-04-18 | 2021-08-26 | ザイリンクス インコーポレイテッドXilinx Incorporated | 異種システムオンチップのためのソフトウェア定義マルチドメイン作成分離 |
JP7250043B2 (ja) | 2018-04-18 | 2023-03-31 | ザイリンクス インコーポレイテッド | 異種システムオンチップのためのソフトウェア定義マルチドメイン作成分離 |
CN108958910A (zh) * | 2018-05-21 | 2018-12-07 | 福建省数字福建云计算运营有限公司 | 一种基于异构环境下的任务调度方法及终端 |
CN108958910B (zh) * | 2018-05-21 | 2020-12-18 | 福建省数字福建云计算运营有限公司 | 一种基于异构环境下的任务调度方法及终端 |
JP7571414B2 (ja) | 2020-08-06 | 2024-10-23 | 日本電気株式会社 | コンテナ実行システム、端末装置、サーバ装置、コンテナ実行環境決定方法及びプログラム |
WO2021151352A1 (zh) * | 2020-10-23 | 2021-08-05 | 平安科技(深圳)有限公司 | 裸金属服务交付方法、装置、计算机设备和存储介质 |
WO2023105670A1 (ja) * | 2021-12-08 | 2023-06-15 | 日本電信電話株式会社 | リソース管理装置及びプログラム |
WO2023105671A1 (ja) * | 2021-12-08 | 2023-06-15 | 日本電信電話株式会社 | 計算機及びプログラム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2017029826A1 (ja) | 2018-05-10 |
US20180239642A1 (en) | 2018-08-23 |
EP3340051A1 (en) | 2018-06-27 |
EP3340051A4 (en) | 2019-04-10 |
JP6438144B2 (ja) | 2018-12-12 |
CN107924340A (zh) | 2018-04-17 |
US10628225B2 (en) | 2020-04-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6438144B2 (ja) | リソース構成システム、リソース構成方法及びリソース構成プログラム | |
EP3762826B1 (en) | Live migration of virtual machines in distributed computing systems | |
CN109791500B (zh) | 虚拟机实例和客户可编程逻辑之间的中间主机集成电路 | |
US10146563B2 (en) | Predictive layer pre-provisioning in container-based virtualization | |
US11573714B2 (en) | Compressibility instrumented dynamic volume provisioning | |
US11182718B2 (en) | Methods and systems to optimize server utilization for a virtual data center | |
US10324754B2 (en) | Managing virtual machine patterns | |
US8825863B2 (en) | Virtual machine placement within a server farm | |
US8904384B2 (en) | Reducing data transfer overhead during live migration of a virtual machine | |
US20180157560A1 (en) | Methods and apparatus for transparent database switching using master-replica high availability setup in relational databases | |
US10346065B2 (en) | Method for performing hot-swap of a storage device in a virtualization environment | |
US20190229983A1 (en) | Methods and systems that provision applications across multiple computer systems | |
US20150256432A1 (en) | Managing resources in a networked computing environment | |
US10936330B2 (en) | Instantaneous boot of virtual machine instances via remote direct memory access | |
EP3738035A1 (en) | Fast instantiation of virtual machines in distributed computing systems | |
US9965334B1 (en) | Systems and methods for virtual machine storage provisioning | |
US10585736B2 (en) | Incremental dump with fast reboot | |
US11960917B2 (en) | Live migration and redundancy for virtualized storage | |
Chen et al. | Towards Fast Deploying the Coordinated Storage, Computing, and Networking Cloud Service: the Ezilla Toolkit |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16836818 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2017535249 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15742765 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2016836818 Country of ref document: EP |