Patent Application
20130173901
Computer systems having multiple processor sockets form the backbone of the high performance computing industry. The installation of multiple processors into a single system provides considerably greater computational power than systems offering only single processors. Processor manufacturers have integrated the multi-processor capability directly into the chip architecture, Intel's QuickPath Interconnect (“QPI”) and AMD's HyperTransport provide just two examples of how chip manufacturers have sought to leverage multi-processor systems.
Chip manufacturers recognize the value and importance of multi-processor systems, and price processors having the capability to interconnect with one or more additional processors accordingly. Often a processor having its multi-processor interconnect capability disabled can be purchased at a discount over the identical processor having its multi-processor interconnect capability enabled. Where a single, interconnect disabled, processor is supplied in a system, a user desiring to expand the system to a multi-processor system is left with the unenviable (and costly) choice of either replacing the entire system with a factory configured multi-processor system, or replacing the existing, interconnect disabled processor with a new, interconnect enabled processor and then adding a second, similar processor. Both options are costly and inconvenient for the user.
Advantages of one or more disclosed embodiments may become apparent upon reading the following detailed description and upon reference to the drawings in which:
High performance computing platforms increasingly make use of multi-processor capable system architecture. The use of multiple processors within a system adds considerable computational horsepower without the additional costs of attempting to construct a similar system using single processor systems. As used herein, the term “processor” can include any computing device capable of executing one or more instruction sets or sequences of instructions. The term “processor” can therefore include central processing units (CPUs) as well as any other processor configured to execute an instruction.
Processor manufacturers understand the importance of multiple processor systems and often offer processors with interconnect pathways. At times, the processor manufacturer disables the interconnect pathway and offers the processor at a significant discount. Users purchasing a multiple processor capable system with only a single factory processor installed may find that the interconnect pathway on the installed processor has been disabled by the processor manufacturer, thereby limiting the user's ability to subsequently upgrade the system to take advantage of the enhanced performance of a multi-processor system.
The ability to partition a computing device into independent computing subsystems provides flexibility, and sometimes a financial advantage, to users whose computing needs evolve over time or are heterogeneous. Computer partitioning is mostly an expensive proposition reserved for high-end, low-volume systems. The cost of such systems is reflective of the use of specialized processors, chipsets and interconnects required to achieve partitioning. Input/output (I/O) or interconnect solutions for instance, generally require the use of duplicate resources to provide independent resource sets when in partitioned mode. Systems having the ability to use high-volume, low cost, non-partitioned aware, components would drastically reduce the cost of a system with partitioning capabilities.
Multi-processor systems may have unique capabilities, such as hard drive expandability or high graphics card power budgets that are often unavailable on single processor systems. Users desiring such capabilities may purchase a multi-processor system but configure the system with only one processor. Such a solution is cost ineffective from a user's perspective, as they will have paid for unused multi-processor support capability including high layer count printed circuit boards, second processor voltage regulation, expanded motherboard and chassis, additional power supply rails, etc. The provision of systems having these untapped capabilities available for use in providing additional independent computers therefore provides a significant economy to the end user.
A multi-processor computer system is provided. The multi-processor computer system can include a plurality of communicatively coupled processors, each coupled to a common motherboard and each associated with a memory. The system can include a boot code executable from at least one of a standard mode and an independent mode. The plurality of communicatively coupled processors can execute one instance of the boot code in standard mode and at least a portion of the plurality of communicatively coupled processors can execute one instance of the boot code in independent mode.
As used herein, the term “motherboard” can refer to any printed circuit board containing one or more integrated circuits and to which other boards may be coupled. An example might include, but is not limited to, the main printed circuit board containing the basic circuits and expansion ports included in a computing device.
A multi-processor computer method is also provided. The method can include entering an independent mode. Within the independent mode, the method can include retrieving a first boot code from a first boot code storage device and retrieving a second boot code from a second boot code storage device. The method can include executing the first boot code on a first group of processors selected from a plurality of processors coupled to a motherboard while contemporaneously executing the second boot code on a second group of processors selected from the plurality of processors coupled to the motherboard.
Another mufti-processor computer system is also provided. The multi-processor computer system can include two communicatively coupled processors coupled to a common motherboard. The system can further include a first boot code and a first memory accessible to a first of the two communicatively coupled processors. The two communicatively coupled processors can be configured to execute one instance of the first boot code when in a standard mode. A first input/output (I/O) controller can be coupled to the two communicatively coupled processors when in the standard mode. The system can further include a partitioning module. The partitioning module can include a second boot code and a second input/output controller that can be coupled to the second processor when in the independent mode. The second of the two communicatively coupled processors can be configured to execute one instance of the second boot code when in the independent mode. The system can further include a user interface to permit a user to reversibly alternate between at least one of the standard mode or the independent mode.
As used herein, the term “communicative coupling”, or a connection by which devices are “communicatively coupled”, is one by which electromagnetic signals, physical communications, logical communications, or combinations thereof may be transmitted and/or received. Devices referred to as being communicatively coupled to each other can be either directly coupled or coupled through an intermediary physical or logical device. For example, devices communicatively coupled to a motherboard can include devices either directly connected to the motherboard, or communicatively coupled to a daughterboard that is, in turn, communicatively coupled to the motherboard. A communicative coupling may include a physical interface, an electrical interface, a data interface, or combinations thereof sufficient to allow intermittent or continuous communication or control between a plurality of devices. For example, two entities can be communicatively coupled by being able to communicate signals to each other directly or through one or more intermediate entities like a processor, operating system, a logic device, software, or other entity.
The plurality of processors 110 can include any number of physically separate or distinct processors communicatively coupled to a common motherboard 120. In at least some embodiments, all or a portion of the plurality of processors 110 can be physically disposed on a separate circuit board (often referred to as a “daughter board”) that is communicatively coupled to the motherboard 120. In at least some embodiments, all or a portion of the plurality of processors 110 can be disposed in sockets or similar receptacles coupled to the motherboard 120. The plurality of processors 110 can include one or more central processing units (CPUs), or any other type of electronic or logical device configured to execute a sequence containing one or more instructions.
In at least some embodiments, at least a portion of the plurality of processors 110 can include a processor-to-processor interconnect 160 enabling coupling or linkage of a processor to at least one other processor thereby forming a multi-processor computing device. These processor-to-processor interconnect 160 can include any number of systems, devices, or any combinations of systems and devices configured to permit the collaborative execution of one or more instruction sets across two or more processors. Example processor-to-processor interconnects 160 can include, but are not limited to the QuickPath Interconnect (“QPI”) offered by Intel® and the HyperTransport offered by AMD®.
In at least some embodiments, the plurality of processors 110 can include one or more processors having a disabled processor-to-processor interconnect feature. Processors having a disabled processor-to-processor interconnect feature are often priced lower, at times significantly lower, than comparable processors having an enabled processor-to-processor interconnect feature. The cost savings of such disabled processors makes their use economically attractive in computing systems that may have multiple processor sockets but have only one installed, on-board processor at the time of delivery to the user. While the use of a disabled processor may be financially attractive to a system manufacturer, such use often places a user desiring to expand such a system at a significant financial penalty—in such instances, the user is left with the choice of replacing the entire computing system with a multi-processor computing system or replacing the disabled processor with an enabled processor followed by adding a new enabled processor.
The boot code 1301 can include one or more instruction sets configured for execution by one or more of the plurality of processors 110 when power is initially supplied to at least a portion of the plurality of processors 110. In some embodiments, at least a portion of the plurality of processors 110 can access the boot code via an input/output controller 1701. For example, the boot code 1301 can be stored in a read-only memory (ROM) location accessible via the I/O controller 1701. In other embodiments, although not shown in
The memory 140 can be any form or type of volatile or non-volatile storage coupled to the processor 110. In at least some embodiments, the memory 140 can be exclusively associated with a specific processor 110, for example memory 1401 can be exclusively associated with processor 1101, memory 1402 can be exclusively associated with processor 1102, and so on. In other embodiments, the memory 140 can be associated with a group of processors selected from the plurality of processors 110. In some embodiments, the memory 140 can be disposed in whole or in part within the processor 110. The memory 140 can include, in whole or in part, a cache, for example a central processing unit (CPU) cache disposed within the CPU itself.
The processor-to-processor interconnect 160 can include any system or device suitable for providing a bidirectional serial/parallel high-bandwidth, low-latency point-to-point link between some or all of the plurality of processors 110. In some embodiments, the processor-to-processor interconnect 160 can include one or more data transfer layers, for example the Intel® QPI processor-to-processor interconnect having up to five layers: a physical layer, a link layer, a routing layer, a transport layer, and a protocol layer. In some embodiments, the processor-to-processor interconnect 160 can include one or more systems or devices incorporated into some or all of the plurality of processors, the motherboard, or both. For example, processors supplied by Intel® and AMD® may have onboard processor-to-processor interconnect systems or devices. Any or all of the processor-to-processor interconnects 160 can be enabled or disabled at the time of manufacture.
The input/output controller 1701 can include any system, device or combination of systems and devices configured to couple one or more of the plurality of processors 110 to at least one input/output (I/O) device. As depicted in
Although the system 200 will be described in detail with regards to a dual processor system, any number of processors could be similarly grouped, partitioned, and provided with exclusive access to necessary system resources such as power, memory, and the like as needed to provide a physically and logically independent computing device within the system 200. Additionally, while the system 200 will be described in detail with regards to a single partition system created using a single partitioning module 210, any number of similar partitioning modules 210 could be used on a system containing three or more processors to provide at least three partitioned, independently bootable processors, each providing a physically and logically independent computing device, all coupled to a common motherboard 120.
The partitioning module 210 can include any number of systems, devices, or combinations of systems and devices necessary to independently boot at least a portion of the plurality of processors 110, for example, one of the two illustrative processors 1101-2 depicted in
As illustratively depicted in
Although only a second boot code 1302 and a second I/O controller 1702 are depicted in
The partitioning module 210 can be a discrete board mounted component or integrated into another board mounted component. The partitioning module 210, in some embodiments, can be a socket-mount device couplable to an open socket coupled to the common motherboard 120. In at least some embodiments, the partitioning module 210 can be a user installable device.
The user interface 220 can provide the system user with the ability to add or remove partitions within the system 200. For example, even though multiple processors 110 may be deployed in system 200, there may be occasions where not booting one or more groups of processors may be advantageous. In such instances, the user, via the user interface 220, can configure new partitions, delete existing partitions, or interrupt the booting of existing partitions within the system. In some embodiments, the user can make the desired changes to the partition structure or booting sequence via the user interface 220 then reboot the system 200 to enable the entered changes.
The detection module 250 can include any number of systems, devices or any combination of systems and devices configured to detect the insertion of one or more partitioning modules 210 within the system 200. In at least some embodiments, the detection module 250 can interrupt one or more processor-to-processor interconnects 160, thereby enabling the booting of at least a portion of the plurality of processors 110 (e.g. the second group of processors) as a physically discrete computing device coupled to a common motherboard 120. In other embodiments, where no processor-to-processor interconnects 160 are present, or where the processor-to-processor interconnects 160 between the plurality of processors has been disabled by the processor manufacturer, the detection logic 250 can ensure that only one boot code 130 and one I/O controller 170 are coupled to each group of processors 110.
Thus, the partitioning module 210, user interface 220, and detection module 230 can work synergistically to create or remove partitions between two or more groups of processors 110 disposed on a common motherboard 120. The partitioning module 210 can provide all or a portion of the resources necessary to provide independent boot capabilities to one or more groups of processors 110. The user interface can provide the user access to the partitioning scheme, permitting the user to easily and conveniently add, delete, or change the partitions between the groups of processors 110. The detection logic 250 can provide a level of assurance that the partitioning communication pathways have either been established (e.g. establishing the coupling between a processor group, an I/O controller 170, and a boot code 130) or broken (e.g. interrupting the processor-to-processor interconnect linking processors in different processor groups).
The absence of a processor-to-processor interconnect 160 does not impact the operation of the system 200, since each processor 110 is allocated the necessary system resources (e.g. boot code 130, I/O controller 170, and memory) required to successfully boot as an independent computing device despite the fact that both processors share a common motherboard 120.
The system can enter an independent mode at 310. Entry into the independent mode can be manual, for example entry based upon system user input into a user interface 220. Entry into the independent mode can also be partially or completely autonomous, for example where detection logic 250 detects the coupling of a partitioning module 210 to the motherboard 120. In either event, a boot code 130 and an I/O controller 170 can be manually or automatically assigned to each of the processor groups. Although not depicted in
After entry into independent mode, a first boot code 1301 can be retrieved at 320. The first boot code 1301 can be associated with a first group of processors 1101 (recall that a processor “group” can contain as few as one processor 110). The first boot code 1301 can be retrieved from a first boot code storage device. The first boot code storage device can be a unique location accessible only by the first processor group 1101. In some embodiments, the first boot code 1301 can be accessed directly by the first group of processors 1101, while in other embodiments, the first boot code 1301 can be accessed via one or more first I/O controllers 1701.
After retrieval at 320, the first boot code 1301 can be executed on a first group of processors 1101 coupled to a motherboard 120 at 330. The execution of the first boot code 1301 on the first group of processors 1101 can provide a first physically isolated, independent computing device within the system.
Contemporaneous with or subsequent to the retrieval of the first boot code 1301 at 320, a second boot code 1302 can be retrieved at 340. The second boot code 1302 can be associated with a second group of processors 1102. The second boot code 1302 can be retrieved from a second boot code storage device. The second boot code storage device can be a unique location accessible only by the second processor group 1102. In some embodiments, the second boot code 1302 can be accessed directly by the second group of processors 1102, while in other embodiments, the second boot code 1302 can be accessed via one or more second I/O controllers 1702.
After retrieval at 340, the second boot code 1302 can be executed on the second group of processors 110 coupled to a motherboard 120 at 350. The first group of processors 1101 and the second group of processors 1102 can be coupled to a common motherboard 120. The execution of the second boot code 1302 on the second group of processors 1102 can provide a second physically isolated, independent computing device within the system.
Although the method described with regard to
The system can enter an independent mode at 410. Entry into the independent mode at 410 can be manual, for example entry based upon system user input into a user interface 220. Entry into the independent mode at 410 can also be partially or completely autonomous, for example where detection logic 250 detects the coupling of a partitioning module 210 to the motherboard 120. In either event, a boot code 130 and an I/O controller 170 can be manually or automatically assigned to each of the processor groups. Although not depicted in
A first I/O controller 1701 can be coupled to a first group of processors 1101 at 420. The first group of processors 1101 can be coupled to a motherboard 120. The first I/O controller 1701 can, among other things, provide the first group of processors 1101 access to a first boot code storage location. In at least some embodiments, the first boot code storage location can be accessible only to the first group of processors 1101. In some embodiments, the first I/O controller 1701 can be coupled to one or more first I/O devices, for example a network interface device such as an Ethernet interface.
The first boot code 1301 can be retrieved by the first group of processors 1101 from the first boot code storage location at 430. In some embodiments, the first boot code 1301 can be accessed directly by the first group of processors 1101, while in other embodiments the first boot code 1301 can be accessed via the first I/O controller 1701.
After retrieval at 430, the first boot code 1301 can be executed by the first group of processors 1101 coupled to a motherboard 120 at 440. The execution of the first boot code 1301 by the first group of processors 1101 can provide a first physically isolated, independent computing device within the system.
At least one first I/O device can be accessed by the first group of processors 1101 via the first I/O controller 1701 at 450. In at least some embodiments, the first I/O device can include one or more network interfaces, for example one or more Ethernet interfaces. In other embodiments, the first I/O device can include one or more communications busses, for example one or more communications busses coupled to additional I/O devices.
Contemporaneous with or subsequent to the coupling of the first I/O controller 1701 to the first group of processors 1101 at 420, a second I/O controller 1702 can be coupled to a second group of processors 1102 at 460. The second group of processors 1102 can be coupled to a common motherboard 120 shared with the first group of processors 1101. The second I/O controller 1702 can, among other things, provide the second group of processors 1102 with access to a second boot code storage location. In at least some embodiments, the second boot code storage location can be accessible only to the second group of processors 1102. In some embodiments, the second I/O controller 1702 can be coupled to one or more second I/O devices, for example a network interface device such as an Ethernet interface.
Contemporaneous with or subsequent to the retrieval of the first boot code 1301 at 430, the second boot code 1302 can be retrieved by the second group of processors 1102 from the second boot code storage location at 470. In some embodiments, the second boot code 1302 can be accessed directly by the second group of processors 1102, while in other embodiments the second boot code 1302 can be accessed via the second I/O controller 1702.
Contemporaneous with or subsequent to the execution of the first boot code 1301 by the first group of processors 1101 at 440, the second boot code 1302 can be executed by the second group of processors 1102 coupled to the motherboard 120 at 480. The execution of the second boot code 1302 by the second group of processors 1102 can provide a second physically isolated, independent computing device within the system.
At least one second I/O device can be accessed by the second group of processors 1102 via the second I/O controller 1702 at 490. In at least some embodiments, the second I/O device can include one or more network interfaces, for example one or more Ethernet interfaces. In other embodiments, the second I/O device can include one or more communications busses, for example one or more communications busses coupled to additional I/O devices.
Although the method 400 described with regard to
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/US10/55021 | 11/1/2010 | WO | 00 | 3/7/2013 |
