WO2016000224A1 - Computer system - Google Patents

Computer system Download PDF

Info

Publication number
WO2016000224A1
WO2016000224A1 PCT/CN2014/081491 CN2014081491W WO2016000224A1 WO 2016000224 A1 WO2016000224 A1 WO 2016000224A1 CN 2014081491 W CN2014081491 W CN 2014081491W WO 2016000224 A1 WO2016000224 A1 WO 2016000224A1
Authority
WO
WIPO (PCT)
Prior art keywords
physical
resource
switching unit
computer system
chip
Prior art date
Application number
PCT/CN2014/081491
Other languages
French (fr)
Chinese (zh)
Inventor
肖世海
杨伟
赵俊峰
陈少杰
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to PCT/CN2014/081491 priority Critical patent/WO2016000224A1/en
Priority to CN201480038267.8A priority patent/CN105393504B/en
Publication of WO2016000224A1 publication Critical patent/WO2016000224A1/en

Links

Definitions

  • Embodiments of the present invention relate to computer technology, and in particular, to a computer system. Background technique
  • processor cores In a conventional multi-core computer system, multiple processor cores share resources such as a cache, a network-on-chip (NoC), and a memory. As the number of processor cores increases, the scale of the on-chip network increases dramatically, and the coherent protocol overhead, power consumption overhead, and latency overhead become larger, which affects the development of multi-core computer systems toward more processor cores. . Therefore, the industry has proposed a new multi-core processor architecture, which is mainly to divide a multi-core processor into multiple physically separate physical partitions, each of which can independently run its own operating system or virtualization management software. .
  • FIG. 1 is a schematic diagram of an architecture of an existing multi-core processor.
  • a multi-core processor is divided into two physical partitions: physical partition 1 and physical partition 2, and physical partitions of physical partition 1 and physical partition 2 are partitioned.
  • the partitioning has been fixed.
  • Physical resources include: kernel, cache, memory, I/O resources, and so on.
  • the two physical partitions are independent.
  • Each physical partition has a separate on-chip network.
  • physical partition 1 and physical partition 2 share only memory and I/O, and do not share the kernel and the last level of cache ( Last Level Cache (LLC) means that the processor core and LLC of the two physical partitions are independent.
  • LLC Last Level Cache
  • C represents the processor core.
  • 2 is a schematic structural diagram of a conventional on-chip network.
  • R is an exchange unit of an on-chip network, and each switching unit is connected and communicated with an adjacent switching unit through an I/O channel, and each exchange is performed.
  • the unit is connected to the physical resource through the mi interface, and the connection between the physical resource and the switching unit in the existing multi-core processor is fixed, that is, the physical resources of each physical partition are fixed.
  • the existing multi-core processor architecture In the existing multi-core processor architecture, the size of the on-chip network of each physical partition is reduced, and the consistency protocol overhead is also small. When the chip needs to integrate more processors, a physical partition can be directly added, thereby The scalability of multi-core processors has also been greatly improved.
  • the existing multi-core processor architecture also has the following problems: The physical resources of each physical partition of the multi-core processor are fixed, because the hardware resources required for different applications running on each physical partition are different, some physical The partition may face waste of resources, and other physical partitions may face insufficient resources. In the prior art, the resources of each physical partition are fixed, and cannot be adjusted according to resources required by the application, so that the physical resource utilization rate of the chip is low.
  • Embodiments of the present invention provide a computer system capable of dynamically configuring physical resources of physical partitions of a computer system, thereby improving utilization of physical resources.
  • a first aspect of the present invention provides a computer system, including: a physical resource, a plurality of on-chip networks, and a configurable interconnection module;
  • the configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and configure a connection relationship between the physical resource and the plurality of on-chip networks, where each on-chip network and the on-chip network
  • the connected physical resources make up a physical partition.
  • each of the on-chip networks includes multiple switching units
  • the configurable interconnection module includes: multiple resource connection lines, multiple Each of the resource connection lines is connected to the physical resource, and each of the exchange unit connection lines is respectively connected to one switching unit of the on-chip network, and the plurality of resource connections are connected.
  • the line and the plurality of switching unit connecting lines are connected to each other in a crisscross manner, and each cross point is provided with a first switch, and the resource management unit controls the opening and closing of the first switch to control the
  • the physical resources connected to the resource connection line at the cross point are connected and disconnected from the exchange unit connected to the exchange unit connection line at the cross point.
  • the configurable interconnection module further includes: a plurality of communication connection lines and a communication management unit, The plurality of communication connecting lines and the plurality of resource connecting lines are connected to each other in a crisscross manner, and each cross point is provided with a second switch, and the communication management unit controls on and off of the second switch
  • the second switch is configured to control communication and disconnection between the exchange unit connected to the corresponding switching unit connection at the cross point and the other switching unit
  • the physical resource includes Any one or combination of resources: multiple processor cores, cache, memory, external memory, input/output I/O, acceleration unit.
  • the resource connection line, the exchange The unit connection line and the communication connection line are optical waveguides
  • the first switch and the second switch are optical switches.
  • the physical resource is divided into configurable a resource and a fixed resource, where the switching unit of each of the on-chip networks is divided into a configurable switching unit and a fixed switching unit, where the configurable resource is connected to the configurable switching unit of the on-chip network through the configurable interconnection module.
  • the fixed resource is connected to a fixed switching unit of the on-chip network, and a connection relationship between the fixed resource and a fixed switching unit of the on-chip network is fixed.
  • the configurable interconnection module further includes:
  • An obtaining unit configured to obtain a performance analysis result of each physical partition, where the performance analysis result includes usage of physical resources of each partition;
  • an adjusting unit configured to adjust physical resources of the physical partitions according to performance analysis results of the physical partitions.
  • the adjusting unit is specifically configured to: use, according to physical resources of each physical partition In the case, the physical resources of the physical partition with low resource utilization are released, and the physical resources of the physical partition with the shortage of physical resources are increased.
  • the computer system of the embodiment of the present invention by adding a configurable interconnection module between the physical resource and the on-chip network, the configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and the flexible configuration through the configurable interconnection module
  • the connection relationship between the physical resources and the plurality of on-chip networks enables the resources of the computer system to be utilized reasonably, and the resource utilization of the entire computer system is improved.
  • 1 is a schematic diagram of an architecture of an existing multi-core processor
  • FIG. 2 is a schematic structural diagram of a conventional on-chip network
  • FIG. 3 is a schematic structural diagram of a computer system according to Embodiment 1 of the present invention.
  • FIG. 4 is a schematic structural diagram of a computer system according to Embodiment 2 of the present invention.
  • FIG. 5 is a schematic structural diagram of a computer system according to Embodiment 3 of the present invention.
  • FIG. 6 is a schematic structural diagram of a computer system according to Embodiment 4 of the present invention.
  • the technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention.
  • the embodiments are a part of the embodiments of the invention, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.
  • the computer system of this embodiment includes: a physical resource, a plurality of on-chip networks, and a configurable interconnection module, and the configurable interconnection module is used for connecting. And the physical resources and the plurality of on-chip networks, and the connection relationship between the physical resources and the plurality of on-chip networks, wherein each of the on-chip network and the physical resources connected to the on-chip network form a physical partition.
  • the physical resource may include any one or a combination of the following resources: multiple processor cores, caches, memory, external memory, input/output (I/O), acceleration Units, physical resources can also include other resources, not listed here, the processor core can have its own private cache, the private cache of each processor core can not be shared.
  • Each on-chip network includes a plurality of switching units, and the switching units have switching and routing functions.
  • the topology of each on-chip network may be the same or different, and each on-chip network may have an independent consistency protocol.
  • each switching unit of the network on the chip passes through a configurable interconnect module.
  • Physical resources are connected.
  • Each switching unit has different physical resources. Some switching units are connected to the processor core. Some switching units are connected to the cache. Some switching units are connected to the memory.
  • Each on-chip network and its connected physical resources form a physical partition. The physical partitions are independent. Therefore, different physical resources can be configured for each physical partition according to actual needs.
  • connection relationship between the physical resource and the on-chip network is not fixed, that is, the size and connection relationship of each physical partition are not fixed, and the connection relationship between the physical resource and the on-chip network can be flexibly configured through the configurable interconnection module.
  • the configurable interconnect module can adjust the physical resources of the two physical partitions to allocate physical resources to physical physical resources. Partition use makes the resources of the computer system be used reasonably and improves the resource utilization of the entire computer system.
  • the configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and the flexible configuration through the configurable interconnection module
  • the connection relationship between the physical resources and the plurality of on-chip networks enables the resources of the computer system to be reasonably utilized, thereby improving resource utilization of the entire computer system.
  • the configurable interconnection module includes: multiple resource connection lines, multiple exchange unit connection lines, and resources.
  • a management unit each of the resource connection lines is respectively connected to the physical resource
  • each of the exchange unit connection lines is respectively connected to one switching unit of the on-chip network
  • the plurality of resource connection lines are exchanged with the plurality of resources
  • the unit connection lines are connected to each other in a crisscross manner, and each cross point is provided with a first switch, and the resource management unit controls the resource connection at the cross point by controlling the on and off of the first switch.
  • the physical resources connected to the line are connected and disconnected from the switching unit to which the switching unit connection line at the cross point is connected.
  • physical resources such as a processor core, a cache, a memory, and an I/O are respectively connected to a resource connection line, and each switching unit of the on-chip network is connected to a switching unit connection line, and when the resource connection line and the exchange unit are connected.
  • the first switch on the line intersection is turned on, it indicates that the physical resource connected to the resource connection line is in communication with the exchange unit connected to the exchange unit connection line.
  • the first switch is disconnected, it indicates that the resource connection line is connected.
  • the physical resource is disconnected from the switching unit connected to the switching unit connection line, and the on/off of the first switch is controlled by the resource management unit, and the resource management unit controls each first open Turn off the switch to achieve flexible configuration of physical resources.
  • the optical waveguide and the switching unit connection line can all adopt optical waveguides.
  • the optical waveguide can meet the requirements of the communication delay within the computer system because of the large bandwidth and high transmission rate.
  • other transmission lines may be used for the resource connection line and the switching unit, which is not limited by the embodiment of the present invention.
  • the first switch can be implemented by using an optical switch, such as a micro-ring optical switch, and the micro-ring optical switch can control the optical coupling capability through a voltage. When a specific voltage is applied to the micro-ring optical switch, the micro-ring optical switch takes the optical signal from the optical signal.
  • One optical waveguide is coupled to the other optical waveguide, while the micro-ring optical switch does not have optical coupling capability at other voltages.
  • the optical waveguide has a wavelength division multiplexing characteristic, that is, one optical waveguide can transmit signals of multiple wavelengths. Therefore, in this embodiment, multiple resource connection lines can use one optical waveguide, and multiple switching unit connection lines can also use one optical path. waveguide. When one optical waveguide can only transmit one wavelength signal, then each resource connection line corresponds to one optical waveguide, and each switching unit connection line corresponds to one optical waveguide.
  • optical switches can use wide-spectrum optical switches, and wide-spectrum optical switches can couple optical signals of multiple wavelengths at once.
  • FIG. 5 is a schematic structural diagram of a computer system according to Embodiment 3 of the present invention.
  • the computer system of this embodiment is further based on the computer system shown in FIG. 4, and the configurable interconnection module further includes: Multiple communication cables and communication management units.
  • the plurality of communication connection lines and the plurality of resource connection lines are connected to each other in a crisscross manner, and each cross point is provided with a second switch, and the second switch is used to control the cross point
  • the corresponding switching unit is connected to the connection and disconnection between the connected switching unit and the other switching unit, and the communication management unit is configured to control the switching of the second switch.
  • each switching unit on the network on the chip can only communicate directly with the adjacent switching unit. If a switching unit needs to communicate with a switching unit that is not adjacent to it, the forwarding needs to be performed through the forwarding of the intermediate switching unit. And in the prior art, the switching units of the two on-chip networks are not able to communicate.
  • the solution of the embodiment of the present invention can directly interconnect two switching units that are not directly connected by using a communication connection line.
  • the two switching units that are not directly connected here may be two switching units on the same on-chip network, or may be located at Two switching units of different on-chip networks.
  • the configurable interconnection module interconnects two switching units that are not directly interconnected on the same on-chip network, enabling direct communication between any two switching units on the same on-chip network, reducing the delay of data packet transmission in the on-chip network.
  • Configurable interconnect modules configure two exchange orders on different on-chip networks Meta-interconnection allows direct communication between different physical partitions.
  • the optical waveguide, the switching unit connection line, and the communication connection line may each adopt an optical waveguide, and the first switch and the second switch may both adopt a micro-ring optical switch.
  • FIG. 6 is a schematic structural diagram of a computer system according to Embodiment 4 of the present invention.
  • the difference between the embodiment and the embodiment shown in FIG. 4 and FIG. 5 is that the connection relationship between the physical resource and the on-chip network is only partially configurable, and part of the physical resources and The connection relationship between the switching units of the on-chip network is fixed.
  • the physical resource is divided into a configurable resource and a fixed resource
  • the switching unit of each of the on-chip networks is divided into a configurable switching unit and a fixed switching unit
  • the configurable resource passes the configurable interconnecting module.
  • the fixed resource Connected to the configurable switching unit of the on-chip network, the fixed resource is connected to the fixed switching unit of the on-chip network, and the connection relationship between the fixed resource and the fixed switching unit of the on-chip network is fixed.
  • connection relationship between the configurable resource and the configurable switching unit is not fixed, and the configurable interconnection module can be flexibly configured, and the connection relationship between the fixed resource and the fixed switching unit is fixed.
  • the computing power required for the physical partition is basically stable, and the storage resources required for the physical partition are highly volatile, so the processor core can be used as a fixed resource, and the cache,
  • a storage resource such as a memory is a configurable resource, and the processor core is directly connected to a fixed switching unit of the on-chip network, and the cache and the memory are connected to the configurable switching unit of the on-chip network through the configurable interconnection module.
  • the configurable interconnection module further includes: an obtaining unit, configured to obtain performance analysis results of the physical partitions, where the performance analysis result includes the The usage of the physical resources of each partition; the adjusting unit, configured to adjust physical resources of the physical partitions according to performance analysis results of the physical partitions.
  • Each physical partition can run a separate operating system or virtualization management software.
  • the performance of the respective physical partitions may be analyzed by the operating system or the virtualization management software, and the obtaining unit may receive performance analysis results of the physical partitions sent by the operating system or the virtualization management software.
  • the performance analysis results mainly include the usage of physical resources of each physical partition, such as the usage rate of the processor core, the utilization of the cache and the memory, the usage of the 10, and the traffic of the network on the chip.
  • the adjustment unit is connected to the resource management unit and the communication management unit, and the adjustment unit may notify the resource management unit and the communication management unit to reconfigure the physical resources of each physical partition. When a physical partition has excess physical resources or insufficient physical resources, the adjustment unit notifies the resource management unit and the communication management unit The physical resources of the physical partition and the connection relationship of the switching unit are changed, and the physical resources of the physical partition are dynamically adjusted.
  • the adjusting unit determines whether each physical partition has a low physical resource utilization rate or a shortage of physical resources according to the usage of physical resources of each physical partition.
  • the adjusting unit Consider the configuration of each physical partition and decide whether to adjust the configuration.
  • the factors considered by the adjustment unit mainly include: Which physical resources should be released for physical partitions with low resource utilization, or if the physical partition is in response to an unexpected situation, consider maintaining the status quo and not adjusting the physical resources of the physical partition.
  • the adjustment unit should consider the physical resource configuration of all physical partitions of the computer system and decide whether to adjust the configuration of physical resources.
  • the adjustment unit decides to reallocate physical resources, it needs to perform a reallocation handshake with the operating system or virtualization management software of each physical partition. After the reassignment handshake is completed, the operating system or the virtualization management software prepares for the switching of the physical resources, and the adjustment unit notifies the resource management unit and the communication management unit to perform physical resource reallocation.
  • the adjusting unit notifies the operating system or the virtualization management software of each physical partition, which physical resources of the physical partition need to be reallocated, and when the operating system or the virtualization management software receives the physical resource reallocation notification of the adjusting unit, if necessary Switching the processor core from the current physical partition to another physical partition, the operating system or virtualization management software migrates all data on the processor core to other processor cores of the current physical partition; if the cache needs to be moved from the current physical partition To switch to another physical partition, you need to write the dirty data in the cache back to the memory. If you need to switch the memory from the current physical partition to another physical partition, you need to migrate the valid data in the memory.

Landscapes

  • Multi Processors (AREA)

Abstract

Provided is a computer system. The computer system comprises a physical resource, multiple networks-on-chip, and a configurable interconnection module. The configurable interconnection module is used for connecting the physical resource to the multiple networks-on-chip, and for configuring a connection relationship between the physical resource and the multiple networks-on-chip. Each network-on-chip and the physical resource connected to the network-on-chip form a physical partition. By adding the configurable interconnection module between the physical resource and the networks-on-chip, the connection relationship between the physical resource and the multiple networks-on-chip can be flexibly configured by means of the configurable interconnection module, resources of the computer system are properly used, and the resource utilization rate of the whole computer system is improved.

Description

计算机系统  computer system
技术领域 Technical field
本发明实施例涉及计算机技术, 尤其涉及一种计算机系统。 背景技术  Embodiments of the present invention relate to computer technology, and in particular, to a computer system. Background technique
传统的多核计算机系统, 其多个处理器核共享缓存 (cache) 、 片上网络 (Network-on-Chip, 简称 NoC) 、 内存 (Memory) 等资源。 随着处理器核的 数量增加, 片上网络的规模急剧增大, 一致性协议开销、 功耗开销、 延时开 销也越来越大, 从而影响了多核计算机系统向更多处理器核的方向发展。 因 此, 业界提出了一种新型的多核处理器架构, 该架构主要是将多核处理器划 分为多个物理上独立的物理分区, 每个物理分区可以独立的运行自己的操作 系统或者虚拟化管理软件。  In a conventional multi-core computer system, multiple processor cores share resources such as a cache, a network-on-chip (NoC), and a memory. As the number of processor cores increases, the scale of the on-chip network increases dramatically, and the coherent protocol overhead, power consumption overhead, and latency overhead become larger, which affects the development of multi-core computer systems toward more processor cores. . Therefore, the industry has proposed a new multi-core processor architecture, which is mainly to divide a multi-core processor into multiple physically separate physical partitions, each of which can independently run its own operating system or virtualization management software. .
图 1为现有的多核处理器的架构示意图, 如图 1所示, 多核处理器被划 分为两个物理分区: 物理分区 1和物理分区 2, 物理分区 1和物理分区 2的物 理资源在分区划分时已经固定, 物理资源包括: 内核、 缓存、 内存、 I/O 资源等。两个物理分区之间是独立的,每个物理分区具有单独的片上网络, 如图 1所示, 物理分区 1和物理分区 2只共享内存和 I/O, 不共享内核和最后 一级缓存 (Last Level Cache, 简称 LLC) , 也就是说两个物理分区的处理 器核和 LLC是独立的, 图中 C表示处理器核。 图 2为一种现有的片上网络的 结构示意图, 如图 2所示, R为片上网络的交换单元, 每个交换单元通过 I/O 通道与相邻的交换单元连接并通信,每个交换单元通过 mi接口与物理资源 连接, 且现有的多核处理器中物理资源与交换单元的连接是固定的, 也就 是说各物理分区的物理资源是固定的。  FIG. 1 is a schematic diagram of an architecture of an existing multi-core processor. As shown in FIG. 1, a multi-core processor is divided into two physical partitions: physical partition 1 and physical partition 2, and physical partitions of physical partition 1 and physical partition 2 are partitioned. The partitioning has been fixed. Physical resources include: kernel, cache, memory, I/O resources, and so on. The two physical partitions are independent. Each physical partition has a separate on-chip network. As shown in Figure 1, physical partition 1 and physical partition 2 share only memory and I/O, and do not share the kernel and the last level of cache ( Last Level Cache (LLC) means that the processor core and LLC of the two physical partitions are independent. In the figure, C represents the processor core. 2 is a schematic structural diagram of a conventional on-chip network. As shown in FIG. 2, R is an exchange unit of an on-chip network, and each switching unit is connected and communicated with an adjacent switching unit through an I/O channel, and each exchange is performed. The unit is connected to the physical resource through the mi interface, and the connection between the physical resource and the switching unit in the existing multi-core processor is fixed, that is, the physical resources of each physical partition are fixed.
现有的多核处理器架构, 每个物理分区的片上网络的规模减小了, 一致 性协议开销也小了, 当芯片需要集成更多处理器时, 可以直接增加一个物理 分区即可, 从而使得多核处理器的扩展性也大大提高了。 但是, 现有的多核 处理器的架构也存在以下问题: 多核处理器的每个物理分区的物理资源是固 定的, 由于每个物理分区上运行的不同应用所需的硬件资源不同, 有些物理 分区可能面临资源浪费, 而其他物理分区可能面临资源不够, 而现有技术中 各物理分区的资源固定, 不能根据应用所需的资源进行调整, 使得芯片的物 理资源利用率低。 发明内容 本发明实施例提供一种计算机系统, 能够动态的配置计算机系统的各物 理分区的物理资源, 提高了物理资源的利用率。 In the existing multi-core processor architecture, the size of the on-chip network of each physical partition is reduced, and the consistency protocol overhead is also small. When the chip needs to integrate more processors, a physical partition can be directly added, thereby The scalability of multi-core processors has also been greatly improved. However, the existing multi-core processor architecture also has the following problems: The physical resources of each physical partition of the multi-core processor are fixed, because the hardware resources required for different applications running on each physical partition are different, some physical The partition may face waste of resources, and other physical partitions may face insufficient resources. In the prior art, the resources of each physical partition are fixed, and cannot be adjusted according to resources required by the application, so that the physical resource utilization rate of the chip is low. SUMMARY OF THE INVENTION Embodiments of the present invention provide a computer system capable of dynamically configuring physical resources of physical partitions of a computer system, thereby improving utilization of physical resources.
本发明第一方面提供一种计算机系统, 包括: 物理资源、 多个片上网络、 可配置互联模块;  A first aspect of the present invention provides a computer system, including: a physical resource, a plurality of on-chip networks, and a configurable interconnection module;
所述可配置互联模块, 用于连接所述物理资源和所述多个片上网络, 以 及配置所述物理资源与所述多个片上网络的连接关系, 所述每个片上网络与 所述片上网络连接的物理资源组成一个物理分区。  The configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and configure a connection relationship between the physical resource and the plurality of on-chip networks, where each on-chip network and the on-chip network The connected physical resources make up a physical partition.
结合本发明第一方面, 在本发明第一方面的第一种可能的实现方式中, 所述每个片上网络包括多个交换单元, 所述可配置互联模块包括: 多条资源 连接线、 多条交换单元连接线、 资源管理单元, 所述每条资源连接线分别 与所述物理资源连接, 所述每条交换单元连接线分别连接所述片上网络的 一个交换单元, 所述多条资源连接线与所述多条交换单元连接线以十字交 叉的方式互相连接, 每个十字交叉点上设置有第一开关, 所述资源管理单 元通过控制所述第一开关的通断, 以控制所述十字交叉点上的资源连接线 所连接的物理资源与所述十字交叉点上的交换单元连接线所连接的交换 单元的连通和断开。  With reference to the first aspect of the present invention, in a first possible implementation manner of the first aspect of the present disclosure, each of the on-chip networks includes multiple switching units, where the configurable interconnection module includes: multiple resource connection lines, multiple Each of the resource connection lines is connected to the physical resource, and each of the exchange unit connection lines is respectively connected to one switching unit of the on-chip network, and the plurality of resource connections are connected. The line and the plurality of switching unit connecting lines are connected to each other in a crisscross manner, and each cross point is provided with a first switch, and the resource management unit controls the opening and closing of the first switch to control the The physical resources connected to the resource connection line at the cross point are connected and disconnected from the exchange unit connected to the exchange unit connection line at the cross point.
结合本发明第一方面的第一种可能的实现方式, 在本发明第一方面的第 二种可能的实现方式中, 所述可配置互联模块还包括: 多条通信连接线和 通信管理单元, 所述多条通信连接线与所述多条资源连接线以十字交叉的 方式互相连接, 每个十字交叉点上设置有第二开关, 所述通信管理单元通 过控制所述第二开关的通断, 以所述第二开关用于控制所述十字交叉点上 对应的交换单元连接上连接的交换单元与其他交换单元之间的连通和断 开,  With reference to the first possible implementation manner of the first aspect of the present invention, in a second possible implementation manner of the first aspect of the present invention, the configurable interconnection module further includes: a plurality of communication connection lines and a communication management unit, The plurality of communication connecting lines and the plurality of resource connecting lines are connected to each other in a crisscross manner, and each cross point is provided with a second switch, and the communication management unit controls on and off of the second switch The second switch is configured to control communication and disconnection between the exchange unit connected to the corresponding switching unit connection at the cross point and the other switching unit,
结合本发明第一方面以及本发明第一方面的第一种和第二种可能的实现 方式, 在本发明第一方面的第三种可能的实现方式中, 所述物理资源包括以 下资源中的任意一个或其组合: 多个处理器核、 缓存、 内存、 外存、 输入 /输 出 I/O、 加速单元。 With reference to the first aspect of the present invention and the first and second possible implementation manners of the first aspect of the present invention, in a third possible implementation manner of the first aspect of the present invention, the physical resource includes Any one or combination of resources: multiple processor cores, cache, memory, external memory, input/output I/O, acceleration unit.
结合本发明第一方面以及本发明第一方面的第一种和第二种可能的实现 方式, 在本发明第一方面的第四种可能的实现方式中, 所述资源连接线、 所 述交换单元连接线、 所述通信连接线均为光波导, 所述第一开关和所述第 二开关均为光开关。  With reference to the first aspect of the present invention and the first and second possible implementation manners of the first aspect of the present invention, in a fourth possible implementation manner of the first aspect of the present invention, the resource connection line, the exchange The unit connection line and the communication connection line are optical waveguides, and the first switch and the second switch are optical switches.
结合本发明第一方面以及本发明第一方面的第一种至第四种可能的实现 方式, 在本发明第一方面的第五种可能的实现方式中, 所述物理资源被划分 为可配置资源和固定资源, 所述每个片上网络的交换单元被划分为可配置 交换单元和固定交换单元, 所述可配置资源通过所述可配置互联模块与所 述片上网络的可配置交换单元连接, 所述固定资源与所述片上网络的固定交 换单元连接,所述固定资源与所述片上网络的固定交换单元的连接关系固定。  With reference to the first aspect of the present invention and the first to fourth possible implementation manners of the first aspect of the present invention, in a fifth possible implementation manner of the first aspect of the present disclosure, the physical resource is divided into configurable a resource and a fixed resource, where the switching unit of each of the on-chip networks is divided into a configurable switching unit and a fixed switching unit, where the configurable resource is connected to the configurable switching unit of the on-chip network through the configurable interconnection module. The fixed resource is connected to a fixed switching unit of the on-chip network, and a connection relationship between the fixed resource and a fixed switching unit of the on-chip network is fixed.
结合本发明第一方面以及本发明第一方面的第一种至第五种可能的实现 方式, 在本发明第一方面的第六种可能的实现方式中, 所述可配置互联模块 还包括:  With reference to the first aspect of the present invention and the first to fifth possible implementation manners of the first aspect of the present invention, in a sixth possible implementation manner of the first aspect, the configurable interconnection module further includes:
获取单元, 用于获取所述各物理分区的性能分析结果, 所述性能分析 结果包括所述各分区的物理资源的使用情况;  An obtaining unit, configured to obtain a performance analysis result of each physical partition, where the performance analysis result includes usage of physical resources of each partition;
调整单元, 用于根据所述各物理分区的性能分析结果调整所述各物理 分区的物理资源。  And an adjusting unit, configured to adjust physical resources of the physical partitions according to performance analysis results of the physical partitions.
结合本发明第一方面的第六种可能的实现方式, 在本发明第一方面的第 七种可能的实现方式中, 所述调整单元具体用于: 根据所述各物理分区的 物理资源的使用情况, 释放资源利用率低的物理分区的物理资源, 增加物 理资源短缺的物理分区的物理资源。  With reference to the sixth possible implementation manner of the first aspect of the present invention, in a seventh possible implementation manner of the first aspect of the present disclosure, the adjusting unit is specifically configured to: use, according to physical resources of each physical partition In the case, the physical resources of the physical partition with low resource utilization are released, and the physical resources of the physical partition with the shortage of physical resources are increased.
本发明实施例的计算机系统, 通过在物理资源和片上网络之间增加可配 置互联模块, 可配置互联模块用于连接所述物理资源和所述多个片上网络, 通过可配置互联模块灵活的配置所述物理资源与所述多个片上网络的连接关 系, 使计算机系统的资源得到合理的利用, 提高整个计算机系统的资源利用  The computer system of the embodiment of the present invention, by adding a configurable interconnection module between the physical resource and the on-chip network, the configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and the flexible configuration through the configurable interconnection module The connection relationship between the physical resources and the plurality of on-chip networks enables the resources of the computer system to be utilized reasonably, and the resource utilization of the entire computer system is improved.
附图说明 为了更清楚地说明本发明实施例或现有技术中的技术方案, 下面将对 实施例或现有技术描述中所需要使用的附图作一简单地介绍, 显而易见 地, 下面描述中的附图是本发明的一些实施例, 对于本领域普通技术人员 来讲, 在不付出创造性劳动性的前提下, 还可以根据这些附图获得其他的 附图。 DRAWINGS In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief description of the drawings used in the embodiments or the prior art description will be briefly described below. Obviously, the drawings in the following description It is a certain embodiment of the present invention, and other drawings can be obtained from those skilled in the art without any inventive labor.
图 1为现有的多核处理器的架构示意图;  1 is a schematic diagram of an architecture of an existing multi-core processor;
图 2为一种现有的片上网络的结构示意图;  2 is a schematic structural diagram of a conventional on-chip network;
图 3为本发明实施例一提供的计算机系统的结构示意图;  3 is a schematic structural diagram of a computer system according to Embodiment 1 of the present invention;
图 4为本发明实施例二提供的计算机系统的结构示意图;  4 is a schematic structural diagram of a computer system according to Embodiment 2 of the present invention;
图 5为本发明实施例三提供的计算机系统的结构示意图;  FIG. 5 is a schematic structural diagram of a computer system according to Embodiment 3 of the present invention; FIG.
图 6为本发明实施例四提供的计算机系统的结构示意图。 具体实施方式 为使本发明实施例的目的、 技术方案和优点更加清楚, 下面将结合本 发明实施例中的附图, 对本发明实施例中的技术方案进行清楚、 完整地描 述, 显然,所描述的实施例是本发明一部分实施例, 而不是全部的实施例。 基于本发明中的实施例, 本领域普通技术人员在没有作出创造性劳动前提 下所获得的所有其他实施例, 都属于本发明保护的范围。  FIG. 6 is a schematic structural diagram of a computer system according to Embodiment 4 of the present invention. The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. The embodiments are a part of the embodiments of the invention, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.
图 3为本发明实施例一提供的计算机系统的结构示意图,如图 3所示, 本实施例的计算机系统包括:物理资源、多个片上网络和可配置互联模块, 可配置互联模块用于连接所述物理资源和所述多个片上网络, 以及配置所述 物理资源与所述多个片上网络的连接关系, 所述每个片上网络与所述片上网 络连接的物理资源组成一个物理分区。  3 is a schematic structural diagram of a computer system according to Embodiment 1 of the present invention. As shown in FIG. 3, the computer system of this embodiment includes: a physical resource, a plurality of on-chip networks, and a configurable interconnection module, and the configurable interconnection module is used for connecting. And the physical resources and the plurality of on-chip networks, and the connection relationship between the physical resources and the plurality of on-chip networks, wherein each of the on-chip network and the physical resources connected to the on-chip network form a physical partition.
其中, 物理资源可以包括以下资源中的任意一个或其组合: 多个处理器 核、 缓存 (cache) 、 内存 (memory)、 外存、 输入 /输出 (Input/Output, 简称 I/O) 、 加速单元, 物理资源还可以包括其他资源, 这里不再一一列举, 处理 器核可以带有自己的私有 cache, 各处理器核的私有 cache不能共享。  The physical resource may include any one or a combination of the following resources: multiple processor cores, caches, memory, external memory, input/output (I/O), acceleration Units, physical resources can also include other resources, not listed here, the processor core can have its own private cache, the private cache of each processor core can not be shared.
每个片上网络包括多个交换单元, 交换单元具有交换和路由功能, 每个 片上网络的拓扑结构可以相同也可以不同,每个片上网络可以具有独立的一 致性协议。 如图 3所示, 片上网络的每个交换单元通过可配置互联模块与 物理资源连接, 每个交换单元连接的物理资源不同, 有些交换单元连接处 理器核, 有些交换单元连接缓存, 有些交换单元连接内存, 每个片上网络 与其连接的物理资源组成了一个物理分区, 每个物理分区独立, 因此, 可 以根据实际的需要为每个物理分区配置不同的物理资源。 Each on-chip network includes a plurality of switching units, and the switching units have switching and routing functions. The topology of each on-chip network may be the same or different, and each on-chip network may have an independent consistency protocol. As shown in Figure 3, each switching unit of the network on the chip passes through a configurable interconnect module. Physical resources are connected. Each switching unit has different physical resources. Some switching units are connected to the processor core. Some switching units are connected to the cache. Some switching units are connected to the memory. Each on-chip network and its connected physical resources form a physical partition. The physical partitions are independent. Therefore, different physical resources can be configured for each physical partition according to actual needs.
本实施例的计算机系统中, 物理资源和片上网络的连接关系不固定, 也 就是说每个物理分区的大小、 连接关系不固定, 可以通过可配置互联模块灵 活配置物理资源与片上网络的连接关系。 例如, 当某一物理分区的物理资源 空闲时, 而另一些物理分区的物理资源紧张时, 可配置互联模块通过调整这 两个物理分区的物理资源, 将空闲物理资源配置给物理资源紧张的物理分区 使用, 使计算机系统的资源得到合理的利用, 提高整个计算机系统的资源利 用率。  In the computer system of this embodiment, the connection relationship between the physical resource and the on-chip network is not fixed, that is, the size and connection relationship of each physical partition are not fixed, and the connection relationship between the physical resource and the on-chip network can be flexibly configured through the configurable interconnection module. . For example, when the physical resources of a physical partition are idle and the physical resources of other physical partitions are tight, the configurable interconnect module can adjust the physical resources of the two physical partitions to allocate physical resources to physical physical resources. Partition use makes the resources of the computer system be used reasonably and improves the resource utilization of the entire computer system.
本实施例的计算机系统, 通过在物理资源和片上网络之间增加可配置互 联模块, 可配置互联模块用于连接所述物理资源和所述多个片上网络, 通过 可配置互联模块灵活的配置所述物理资源与所述多个片上网络的连接关系, 使计算机系统的资源得到合理的利用, 提高整个计算机系统的资源利用率。  In the computer system of this embodiment, by adding a configurable interconnection module between the physical resource and the on-chip network, the configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and the flexible configuration through the configurable interconnection module The connection relationship between the physical resources and the plurality of on-chip networks enables the resources of the computer system to be reasonably utilized, thereby improving resource utilization of the entire computer system.
图 4为本发明实施例二提供的计算机系统的结构示意图, 如图 4所示, 本实施例的计算机系统中, 可配置互联模块包括: 多条资源连接线、 多条交 换单元连接线、 资源管理单元, 所述每条资源连接线分别与所述物理资源 连接, 所述每条交换单元连接线分别连接所述片上网络的一个交换单元, 所述多条资源连接线与所述多条交换单元连接线以十字交叉的方式互相 连接, 每个十字交叉点上设置有第一开关, 所述资源管理单元通过控制所 述第一开关的通断, 以控制所述十字交叉点上的资源连接线所连接的物理 资源与所述十字交叉点上的交换单元连接线所连接的交换单元的连通和 断开。  4 is a schematic structural diagram of a computer system according to Embodiment 2 of the present invention. As shown in FIG. 4, in the computer system of this embodiment, the configurable interconnection module includes: multiple resource connection lines, multiple exchange unit connection lines, and resources. a management unit, each of the resource connection lines is respectively connected to the physical resource, and each of the exchange unit connection lines is respectively connected to one switching unit of the on-chip network, and the plurality of resource connection lines are exchanged with the plurality of resources The unit connection lines are connected to each other in a crisscross manner, and each cross point is provided with a first switch, and the resource management unit controls the resource connection at the cross point by controlling the on and off of the first switch. The physical resources connected to the line are connected and disconnected from the switching unit to which the switching unit connection line at the cross point is connected.
如图 4所示, 处理器核、 缓存、 内存、 I/O等物理资源分别和资源连接线 连接, 每个片上网络的交换单元分别和交换单元连接线连接, 当资源连接线 和交换单元连接线交叉点上的第一开关导通时, 表示该资源连接线所连接的 物理资源与该交换单元连接线所连接的交换单元连通, 当第一开关断开, 表 示该资源连接线所连接的物理资源与该交换单元连接线所连接的交换单元断 开, 第一开关的通断由资源管理单元控制, 资源管理单元通过控制各第一开 关的通断, 实现物理资源的灵活配置。 As shown in FIG. 4, physical resources such as a processor core, a cache, a memory, and an I/O are respectively connected to a resource connection line, and each switching unit of the on-chip network is connected to a switching unit connection line, and when the resource connection line and the exchange unit are connected. When the first switch on the line intersection is turned on, it indicates that the physical resource connected to the resource connection line is in communication with the exchange unit connected to the exchange unit connection line. When the first switch is disconnected, it indicates that the resource connection line is connected. The physical resource is disconnected from the switching unit connected to the switching unit connection line, and the on/off of the first switch is controlled by the resource management unit, and the resource management unit controls each first open Turn off the switch to achieve flexible configuration of physical resources.
本实施例中, 资源连接线和交换单元连接线均可以采用光波导, 光波导 由于带宽大、 传输速率高, 能够很好的满足计算机系统内部对通信时延的要 求。 当然, 资源连接线和交换单元也可以采用其他的传输线, 本发明实施例 并不对此进行限制。 第一开关可以采用光开关实现, 该光开关例如为微环光 开关, 微环光开关可通过电压控制光耦合能力, 当对微环光开关施加特定 电压时, 微环光开关把光信号从一个光波导耦合到另外一个光波导上去, 而在其他电压下微环光开关不具有光耦合能力。  In this embodiment, the optical waveguide and the switching unit connection line can all adopt optical waveguides. The optical waveguide can meet the requirements of the communication delay within the computer system because of the large bandwidth and high transmission rate. Of course, other transmission lines may be used for the resource connection line and the switching unit, which is not limited by the embodiment of the present invention. The first switch can be implemented by using an optical switch, such as a micro-ring optical switch, and the micro-ring optical switch can control the optical coupling capability through a voltage. When a specific voltage is applied to the micro-ring optical switch, the micro-ring optical switch takes the optical signal from the optical signal. One optical waveguide is coupled to the other optical waveguide, while the micro-ring optical switch does not have optical coupling capability at other voltages.
光波导具有波分复用特性, 即一路光波导可以传输多个波长的信号, 因此, 本实施例中, 多条资源连接线可使用一路光波导, 多条交换单元连 接线也可以使用一路光波导。 当一路光波导只能传输一个波长的信号时, 那么每条资源连接线对应一路光波导, 每条交换单元连接线对应一路光波 导。 在实际的应用中, 光开关可以使用宽频谱光开关, 宽频谱光开关可以 一次性耦合多个波长的光信号。  The optical waveguide has a wavelength division multiplexing characteristic, that is, one optical waveguide can transmit signals of multiple wavelengths. Therefore, in this embodiment, multiple resource connection lines can use one optical waveguide, and multiple switching unit connection lines can also use one optical path. waveguide. When one optical waveguide can only transmit one wavelength signal, then each resource connection line corresponds to one optical waveguide, and each switching unit connection line corresponds to one optical waveguide. In practical applications, optical switches can use wide-spectrum optical switches, and wide-spectrum optical switches can couple optical signals of multiple wavelengths at once.
图 5为本发明实施例三提供的计算机系统的结构示意图,如图 5所示, 本实施例的计算机系统在图 4所示计算机系统的基础上, 进一歩地, 可配 置互联模块还包括: 多条通信连接线和通信管理单元。 其中, 所述多条通 信连接线与所述多条资源连接线以十字交叉的方式互相连接, 每个十字交 叉点上设置有第二开关, 所述第二开关用于控制所述十字交叉点上对应的 交换单元连接上连接的交换单元与其他交换单元之间的连通和断开, 所述 通信管理单元用于控制所述第二开关的通断。  FIG. 5 is a schematic structural diagram of a computer system according to Embodiment 3 of the present invention. As shown in FIG. 5, the computer system of this embodiment is further based on the computer system shown in FIG. 4, and the configurable interconnection module further includes: Multiple communication cables and communication management units. The plurality of communication connection lines and the plurality of resource connection lines are connected to each other in a crisscross manner, and each cross point is provided with a second switch, and the second switch is used to control the cross point The corresponding switching unit is connected to the connection and disconnection between the connected switching unit and the other switching unit, and the communication management unit is configured to control the switching of the second switch.
现有技术中, 片上网络上的各个交换单元只能和相邻的交换单元直接通 信, 若某个交换单元要与其不相邻的交换单元通信, 则需要通过中间交换单 元的转发才能实现通信, 并且现有技术中, 两个片上网络的交换单元不能够 通信。 本发明实施例的方案, 通过通信连接线可以将不直接相连的两个交换 单元直接互联, 这里不直接相连的两个交换单元可以是同一个片上网络上的 两个交换单元, 也可以是位于不同的片上网络的两个交换单元。 可配置互联 模块通过配置同一个片上网络上的不直接互联的两个交换单元互联, 使得同 一个片上网络上任意两个交换单元之间能够直接通信, 减少了片上网络内数 据包传输的时延。 可配置互联模块通过配置位于不同片上网络的两个交换单 元互联, 使得不同物理分区之间可以直接通信。 In the prior art, each switching unit on the network on the chip can only communicate directly with the adjacent switching unit. If a switching unit needs to communicate with a switching unit that is not adjacent to it, the forwarding needs to be performed through the forwarding of the intermediate switching unit. And in the prior art, the switching units of the two on-chip networks are not able to communicate. The solution of the embodiment of the present invention can directly interconnect two switching units that are not directly connected by using a communication connection line. The two switching units that are not directly connected here may be two switching units on the same on-chip network, or may be located at Two switching units of different on-chip networks. The configurable interconnection module interconnects two switching units that are not directly interconnected on the same on-chip network, enabling direct communication between any two switching units on the same on-chip network, reducing the delay of data packet transmission in the on-chip network. . Configurable interconnect modules configure two exchange orders on different on-chip networks Meta-interconnection allows direct communication between different physical partitions.
本实施例中, 资源连接线、 交换单元连接线以及通信连接线均可以采用 光波导, 第一开关和第二开关均可以采用微环光开关。  In this embodiment, the optical waveguide, the switching unit connection line, and the communication connection line may each adopt an optical waveguide, and the first switch and the second switch may both adopt a micro-ring optical switch.
图 6为本发明实施例四提供的计算机系统的结构示意图, 本实施例与图 4和图 5所示实施例的区别在于: 物理资源和片上网络的连接关系只有部分 可配置, 部分物理资源和片上网络的交换单元之间的连接关系是固定的。 具 体地, 所述物理资源被划分为可配置资源和固定资源, 所述每个片上网络 的交换单元被划分为可配置交换单元和固定交换单元, 所述可配置资源通 过所述可配置互联模块与所述片上网络的可配置交换单元连接,所述固定资 源与所述片上网络的固定交换单元连接, 所述固定资源与所述片上网络的固 定交换单元的连接关系固定。  FIG. 6 is a schematic structural diagram of a computer system according to Embodiment 4 of the present invention. The difference between the embodiment and the embodiment shown in FIG. 4 and FIG. 5 is that the connection relationship between the physical resource and the on-chip network is only partially configurable, and part of the physical resources and The connection relationship between the switching units of the on-chip network is fixed. Specifically, the physical resource is divided into a configurable resource and a fixed resource, and the switching unit of each of the on-chip networks is divided into a configurable switching unit and a fixed switching unit, and the configurable resource passes the configurable interconnecting module. Connected to the configurable switching unit of the on-chip network, the fixed resource is connected to the fixed switching unit of the on-chip network, and the connection relationship between the fixed resource and the fixed switching unit of the on-chip network is fixed.
本实施例中, 只有可配置资源与可配置交换单元的连接关系不固定, 可 以通过可配置互联模块进行灵活配置, 固定资源和固定交换单元之间的连接 关系固定。 举例来说, 对于某一物理分区, 该物理分区所需的计算能力基本 上处于稳定, 该物理分区所需的存储资源的波动性较大, 那么可以将处理器 核作为固定资源, 将缓存、 内存等存储资源作为可配置资源, 处理器核与该 片上网络的固定交换单元直接连接, 缓存和内存与该片上网络的可配置交换 单元通过可配置互联模块连接。  In this embodiment, only the connection relationship between the configurable resource and the configurable switching unit is not fixed, and the configurable interconnection module can be flexibly configured, and the connection relationship between the fixed resource and the fixed switching unit is fixed. For example, for a physical partition, the computing power required for the physical partition is basically stable, and the storage resources required for the physical partition are highly volatile, so the processor core can be used as a fixed resource, and the cache, A storage resource such as a memory is a configurable resource, and the processor core is directly connected to a fixed switching unit of the on-chip network, and the cache and the memory are connected to the configurable switching unit of the on-chip network through the configurable interconnection module.
在上述实施例一至实施例四的基础上, 本发明实施例五中, 可配置互联 模块还包括: 获取单元, 用于获取所述各物理分区的性能分析结果, 所述 性能分析结果包括所述各分区的物理资源的使用情况; 调整单元, 用于根 据所述各物理分区的性能分析结果调整所述各物理分区的物理资源。  On the basis of the foregoing Embodiment 1 to Embodiment 4, in the fifth embodiment of the present invention, the configurable interconnection module further includes: an obtaining unit, configured to obtain performance analysis results of the physical partitions, where the performance analysis result includes the The usage of the physical resources of each partition; the adjusting unit, configured to adjust physical resources of the physical partitions according to performance analysis results of the physical partitions.
每个物理分区可运行独立的操作系统或者虚拟化管理软件。 可以由操作 系统或者虚拟化管理软件可对各自的物理分区的性能就进行分析, 获取单元 可以接收操作系统或者虚拟化管理软件发送的各物理分区的性能分析结果。 该性能分析结果主要包括各物理分区的物理资源的使用情况, 例如处理器核 的使用率、 缓存和内存的利用率、 10使用情况、 片上网络的流量情况等。 调 整单元与资源管理单元和通信管理单元连接, 调整单元可以通知资源管理单 元和通信管理单元重新配置各物理分区的物理资源。 当一个物理分区出现物 理资源过剩或物理资源不足时, 调整单元通知资源管理单元和通信管理单元 改变物理分区的物理资源和交换单元的连接关系, 动态的调整物理分区的物 理资源。 Each physical partition can run a separate operating system or virtualization management software. The performance of the respective physical partitions may be analyzed by the operating system or the virtualization management software, and the obtaining unit may receive performance analysis results of the physical partitions sent by the operating system or the virtualization management software. The performance analysis results mainly include the usage of physical resources of each physical partition, such as the usage rate of the processor core, the utilization of the cache and the memory, the usage of the 10, and the traffic of the network on the chip. The adjustment unit is connected to the resource management unit and the communication management unit, and the adjustment unit may notify the resource management unit and the communication management unit to reconfigure the physical resources of each physical partition. When a physical partition has excess physical resources or insufficient physical resources, the adjustment unit notifies the resource management unit and the communication management unit The physical resources of the physical partition and the connection relationship of the switching unit are changed, and the physical resources of the physical partition are dynamically adjusted.
具体地, 调整单元根据各物理分区的物理资源的使用情况, 判断各物理 分区是否有物理资源利用率低或者物理资源短缺的问题, 当发生物理资源利 用率或者物理资源短缺的问题时,调整单元综合考虑各物理分区的配置情况, 决定是否调整配置。 调整单元考虑的因素主要包括: 对于资源利用率低的物 理分区, 应该释放哪些物理资源, 或者, 若该物理分区处于应对突发情况, 则考虑保持现状, 不对该物理分区的物理资源进行调整。 对于物理资源短缺 的物理分区, 需要考虑该物理分区的片上网络是否有空闲的交换单元, 以及 计算机系统当前是否有足够的空闲资源, 当发生空闲资源不足时, 是否通知 其他物理分区释放物理资源, 该物理分区的某个交换单元连接的物理资源是 否可以从一种资源换到另外一种资源, 例如从内存切换到处理器核, 该物理 分区是否保持现状不进行物理资源调整。 调整单元应该综合考虑计算机系统 的所有物理分区的物理资源配置情况, 决定是否需要调整物理资源的配置。  Specifically, the adjusting unit determines whether each physical partition has a low physical resource utilization rate or a shortage of physical resources according to the usage of physical resources of each physical partition. When physical resource utilization or physical resource shortage occurs, the adjusting unit Consider the configuration of each physical partition and decide whether to adjust the configuration. The factors considered by the adjustment unit mainly include: Which physical resources should be released for physical partitions with low resource utilization, or if the physical partition is in response to an unexpected situation, consider maintaining the status quo and not adjusting the physical resources of the physical partition. For a physical partition with a shortage of physical resources, it is necessary to consider whether there is an idle switching unit of the physical partition of the physical partition, and whether the computer system currently has enough free resources, and when the idle resources are insufficient, whether to notify other physical partitions to release physical resources. Whether the physical resource connected to a switching unit of the physical partition can be switched from one resource to another, for example, switching from memory to the processor core, and whether the physical partition maintains the status quo without physical resource adjustment. The adjustment unit should consider the physical resource configuration of all physical partitions of the computer system and decide whether to adjust the configuration of physical resources.
当调整单元决定重新分配物理资源时, 需要和各物理分区的操作系统或 者虚拟化管理软件进行重新分配握手。 当重新分配握手完成后, 操作系统或 者虚拟化管理软件准备好物理资源的切换, 调整单元通知资源管理单元和通 信管理单元进行物理资源重新分配。  When the adjustment unit decides to reallocate physical resources, it needs to perform a reallocation handshake with the operating system or virtualization management software of each physical partition. After the reassignment handshake is completed, the operating system or the virtualization management software prepares for the switching of the physical resources, and the adjustment unit notifies the resource management unit and the communication management unit to perform physical resource reallocation.
具体地, 调整单元通知各个物理分区的操作系统或者虚拟化管理软件, 本物理分区的哪些物理资源需要重新分配, 操作系统或者虚拟化管理软件接 收到调整单元的物理资源重新分配通知时, 若需要将处理器核从当前物理分 区切换到其他物理分区, 操作系统或者虚拟化管理软件将该处理器核上的所 有数据迁移到当前物理分区的其他处理器核上; 若需要将缓存从当前物理分 区切换到其他物理分区, 则需要将该缓存中的脏数据 (duty) 写回内存中; 若需要将内存从当前物理分区切换到其他物理分区, 则需要将该内存中的有 效数据进行迁移。  Specifically, the adjusting unit notifies the operating system or the virtualization management software of each physical partition, which physical resources of the physical partition need to be reallocated, and when the operating system or the virtualization management software receives the physical resource reallocation notification of the adjusting unit, if necessary Switching the processor core from the current physical partition to another physical partition, the operating system or virtualization management software migrates all data on the processor core to other processor cores of the current physical partition; if the cache needs to be moved from the current physical partition To switch to another physical partition, you need to write the dirty data in the cache back to the memory. If you need to switch the memory from the current physical partition to another physical partition, you need to migrate the valid data in the memory.
本领域普通技术人员可以理解: 实现上述方法实施例的全部或部分歩骤 可以通过程序指令相关的硬件来完成, 前述的程序可以存储于一计算机可读 取存储介质中, 该程序在执行时, 执行包括上述方法实施例的歩骤; 而前述 的存储介质包括: ROM、 RAM, 磁碟或者光盘等各种可以存储程序代码的介 质。 A person skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, when executed, The foregoing steps include the steps of the foregoing method embodiments; and the foregoing storage medium includes: a ROM, a RAM, a magnetic disk, or an optical disk, and the like, which can store various program codes. Quality.
最后应说明的是: 以上各实施例仅用以说明本发明的技术方案, 而非对 其限制; 尽管参照前述各实施例对本发明进行了详细的说明, 本领域的普通 技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改, 或者对其中部分或者全部技术特征进行等同替换; 而这些修改或者替换, 并 不使相应技术方案的本质脱离本发明各实施例技术方案的范围。  Finally, it should be noted that the above embodiments are only for explaining the technical solutions of the present invention, and are not intended to be limiting thereof; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that The technical solutions described in the foregoing embodiments may be modified, or some or all of the technical features may be equivalently replaced; and the modifications or substitutions do not deviate from the technical solutions of the embodiments of the present invention. range.

Claims

权 利 要 求 书 Claim
1、 一种计算机系统, 其特征在于, 包括: 物理资源、 多个片上网络、 可 配置互联模块; A computer system, comprising: a physical resource, a plurality of on-chip networks, and a configurable interconnection module;
所述可配置互联模块, 用于连接所述物理资源和所述多个片上网络, 以 及配置所述物理资源与所述多个片上网络的连接关系, 所述每个片上网络与 所述片上网络连接的物理资源组成一个物理分区。  The configurable interconnection module is configured to connect the physical resource and the plurality of on-chip networks, and configure a connection relationship between the physical resource and the plurality of on-chip networks, where each on-chip network and the on-chip network The connected physical resources make up a physical partition.
2、 根据权利要求 1所述的计算机系统, 其特征在于, 所述可配置互联模 块包括: 多条资源连接线、 多条交换单元连接线、 资源管理单元, 所述每 条资源连接线分别与所述物理资源连接, 所述每条交换单元连接线分别连 接所述片上网络的一个交换单元, 所述多条资源连接线与所述多条交换单 元连接线以十字交叉的方式互相连接, 每个十字交叉点上设置有第一开 关, 所述资源管理单元通过控制所述第一开关的通断, 以控制所述十字交 叉点上的资源连接线所连接的物理资源与所述十字交叉点上的交换单元 连接线所连接的交换单元的连通和断开。  2. The computer system according to claim 1, wherein the configurable interconnection module comprises: a plurality of resource connection lines, a plurality of switching unit connection lines, and a resource management unit, wherein each of the resource connection lines respectively The physical resource is connected, and each of the switching unit connection lines is respectively connected to one switching unit of the on-chip network, and the plurality of resource connection lines and the plurality of switching unit connection lines are connected to each other in a crisscross manner. a first switch is disposed on the cross point, and the resource management unit controls the physical resource connected to the resource connection line at the cross point and the cross point by controlling the on and off of the first switch The switching unit connected to the upper switching unit is connected and disconnected.
3、 根据权利要求 2所述的计算机系统, 其特征在于, 所述可配置互 联模块还包括: 多条通信连接线和通信管理单元, 所述多条通信连接线与 所述多条资源连接线以十字交叉的方式互相连接, 每个十字交叉点上设置 有第二开关, 所述通信管理单元通过控制所述第二开关的通断, 以所述第 二开关用于控制所述十字交叉点上对应的交换单元连接上连接的交换单 元与其他交换单元之间的连通和断开,  The computer system according to claim 2, wherein the configurable interconnection module further comprises: a plurality of communication connection lines and a communication management unit, the plurality of communication connection lines and the plurality of resource connection lines Connected to each other in a crisscross manner, each cross point is provided with a second switch, the communication management unit controls the opening and closing of the second switch, and the second switch is used to control the cross point The corresponding switching unit is connected to disconnect and disconnect between the connected switching unit and other switching units.
4、 根据权利要求 1-3中任一项所述的计算机系统, 其特征在于, 所述物 理资源包括以下资源中的任意一个或其组合: 多个处理器核、 缓存、 内存、 外存、 输入 /输出 I/O、 加速单元。  The computer system according to any one of claims 1 to 3, wherein the physical resource comprises any one or a combination of the following resources: a plurality of processor cores, a cache, a memory, an external memory, Input/output I/O, acceleration unit.
5、 根据权利要求 1-3中任一项所述的计算机系统, 其特征在于, 所述 资源连接线、 所述交换单元连接线、 所述通信连接线均为光波导, 所述第 一开关和所述第二开关均为光开关。  The computer system according to any one of claims 1 to 3, wherein the resource connection line, the exchange unit connection line, and the communication connection line are optical waveguides, and the first switch And the second switch is an optical switch.
6、 根据权利要求 1-5中任一项所述的计算机系统, 其特征在于, 所述 物理资源被划分为可配置资源和固定资源, 所述每个片上网络的交换单元 被划分为可配置交换单元和固定交换单元, 所述可配置资源通过所述可配 置互联模块与所述片上网络的可配置交换单元连接, 所述固定资源与所述片 上网络的固定交换单元连接, 所述固定资源与所述片上网络的固定交换单元 的连接关系固定。 The computer system according to any one of claims 1 to 5, wherein the physical resource is divided into a configurable resource and a fixed resource, and the switching unit of each of the on-chip networks is divided into configurable a switching unit and a fixed switching unit, where the configurable resource is connected to the configurable switching unit of the network on chip by the configurable interconnection module, the fixed resource and the slice The fixed switching unit of the upper network is connected, and the connection relationship between the fixed resource and the fixed switching unit of the on-chip network is fixed.
7、 根据权利要求 1-6中任一项所述的计算机系统, 其特征在于, 所述 可配置互联模块还包括:  The computer system according to any one of claims 1 to 6, wherein the configurable interconnection module further comprises:
获取单元, 用于获取所述各物理分区的性能分析结果, 所述性能分析 结果包括所述各分区的物理资源的使用情况;  An obtaining unit, configured to obtain a performance analysis result of each physical partition, where the performance analysis result includes usage of physical resources of each partition;
调整单元, 用于根据所述各物理分区的性能分析结果调整所述各物理 分区的物理资源。  And an adjusting unit, configured to adjust physical resources of the physical partitions according to performance analysis results of the physical partitions.
8、 根据权利要求 7所述的计算机系统, 其特征在于, 所述调整单元 具体用于: 根据所述各物理分区的物理资源的使用情况, 释放资源利用率 低的物理分区的物理资源, 增加物理资源短缺的物理分区的物理资源。  The computer system according to claim 7, wherein the adjusting unit is specifically configured to: release physical resources of a physical partition with low resource utilization rate according to usage of the physical resources of the physical partitions, and increase Physical resources of physical partitions where physical resources are scarce.
PCT/CN2014/081491 2014-07-02 2014-07-02 Computer system WO2016000224A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2014/081491 WO2016000224A1 (en) 2014-07-02 2014-07-02 Computer system
CN201480038267.8A CN105393504B (en) 2014-07-02 2014-07-02 Computer system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2014/081491 WO2016000224A1 (en) 2014-07-02 2014-07-02 Computer system

Publications (1)

Publication Number Publication Date
WO2016000224A1 true WO2016000224A1 (en) 2016-01-07

Family

ID=55018303

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/081491 WO2016000224A1 (en) 2014-07-02 2014-07-02 Computer system

Country Status (2)

Country Link
CN (1) CN105393504B (en)
WO (1) WO2016000224A1 (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070075734A1 (en) * 2005-10-03 2007-04-05 Honeywell International Inc. Reconfigurable network on a chip
CN101977152A (en) * 2010-11-12 2011-02-16 复旦大学 High-performance network-on-chip system suitable for reconfiguration
CN102158403A (en) * 2011-03-24 2011-08-17 山东大学 Efficient data stream transmission communication structure suitable for on chip network and operating method thereof
CN102780936A (en) * 2012-07-17 2012-11-14 西安电子科技大学 Optical on-chip network system of non-blocking communication and communication method thereof
CN102799560A (en) * 2012-09-07 2012-11-28 上海交通大学 Dynamic reconfigurable subnetting method and system based on network on chip

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6260068B1 (en) * 1998-06-10 2001-07-10 Compaq Computer Corporation Method and apparatus for migrating resources in a multi-processor computer system
CN102202005B (en) * 2011-07-12 2013-08-14 西安电子科技大学 Reconfigurable network on mating plate and configuration method
CN102413036A (en) * 2011-08-22 2012-04-11 复旦大学 Real-time partially and dynamically reconfigurable system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070075734A1 (en) * 2005-10-03 2007-04-05 Honeywell International Inc. Reconfigurable network on a chip
CN101977152A (en) * 2010-11-12 2011-02-16 复旦大学 High-performance network-on-chip system suitable for reconfiguration
CN102158403A (en) * 2011-03-24 2011-08-17 山东大学 Efficient data stream transmission communication structure suitable for on chip network and operating method thereof
CN102780936A (en) * 2012-07-17 2012-11-14 西安电子科技大学 Optical on-chip network system of non-blocking communication and communication method thereof
CN102799560A (en) * 2012-09-07 2012-11-28 上海交通大学 Dynamic reconfigurable subnetting method and system based on network on chip

Also Published As

Publication number Publication date
CN105393504A (en) 2016-03-09
CN105393504B (en) 2018-09-28

Similar Documents

Publication Publication Date Title
US10713203B2 (en) Dynamic partition of PCIe disk arrays based on software configuration / policy distribution
US9332324B2 (en) Method and apparatus for efficient and transparent network management and application coordination for software defined optical switched data center networks
CN104301391B (en) Multi-area optical network data center resource virtualizes mapping method
TWI534629B (en) Data transmission method and data transmission system
US9043526B2 (en) Versatile lane configuration using a PCIe PIe-8 interface
WO2016107023A1 (en) Cloud server system
US11791926B2 (en) Efficiently interconnecting a plurality of computing nodes to form a circuit-switched network
US20210311800A1 (en) Connecting accelerator resources using a switch
US20220206979A1 (en) Communications for field programmable gate array device
Guo et al. RDON: a rack-scale disaggregated data center network based on a distributed fast optical switch
US9116881B2 (en) Routing switch apparatus, network switch system, and routing switching method
CN105763488B (en) Data center aggregation core switch and backboard thereof
US20240160592A1 (en) Memory disaggregation and reallocation
CN104104736A (en) Cloud server and use method thereof
Zhang et al. Optical switching based small-world data center network
CN104125292A (en) Data processing device, cloud server and use method thereof
WO2016000224A1 (en) Computer system
Shu et al. Programmable OPS/OCS hybrid data centre network
CN115334378B (en) All-optical network system, device and control method for resource pooling
US20230261778A1 (en) Transmission device management device, transmission device management system, transmission device management method and program
WO2023222086A1 (en) Method for establishing optical path, and related device
Shao et al. OeIM: An Optoelectronic Interconnection Middleware for the Exascale Computer
Song et al. SRFabric: A Semi‐Reconfigurable Rack Scale Topology
Wang Survey of recent research issues in data center networking
Hou et al. Adaptive service degradation in converged optical and data center networks

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 201480038267.8

Country of ref document: CN

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14896368

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14896368

Country of ref document: EP

Kind code of ref document: A1