Memory Module for a Data Center Compute Sled

TECHNICAL FIELD

Examples described herein are generally related to data centers and particularly to compute sleds comprising physical compute resources in a data center.

BACKGROUND

Advancements in networking have enabled the rise in pools of configurable computing resources. A pool of configurable computing resources may be formed from a physical infrastructure including disaggregate physical resources, for example, as found in large data centers. The physical infrastructure can include a number of resources having processors, memory, storage, networking, power, cooling, etc. Management entities of these data centers can aggregate a selection of the resources to form servers and/or computing hosts. These hosts can subsequently be allocated to execute and/or host system SW (e.g., OSs, VMs, Containers, Applications, or the like). The physical resources include processors, which are often housed along with memory on a single sled. The present disclosure is directed to such sleds comprising processors and memory.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a first example data center.

FIG. 2 illustrates a first example rack of a data center.

FIG. 3 illustrates a second example rack.

FIG. 4 illustrates a third example rack.

FIG. 5 illustrates a first example sled.

FIG. 6 illustrates a second example sled.

FIG. 7 illustrates a third example sled.

FIG. 8A-8C illustrates portions of a fourth example sled.

FIG. 9 illustrates a fifth example sled.

FIG. 10 illustrates a second example data center.

DETAILED DESCRIPTION

Data centers may generally be composed of a large number of racks that can contain numerous types of hardware or configurable resources (e.g., processing units, memory, storage, accelerators, networking, fans/cooling modules, power units, etc.). The types of hardware or configurable resources deployed in data centers may also be referred to as physical resources or disaggregate elements. It is to be appreciated, that the size and number of physical resources within a data center can be large, for example, on the order of hundreds of thousands of physical resources. Furthermore, these physical resources can be pooled to form virtual computing platforms for a large number and variety of computing tasks.

These physical resources are often arranged in racks within a data center. The present disclosure provides racks arranged to receive a number of sleds, where each sled can house a number of physical resources. Some of the sleds in a data center can house processor components, such as, central processing units (CPUs), or the like. Such processing components are typically paired with memory resources. For example, a CPU can be paired with memory to facilitate operations (e.g., executing instructions, performing processing operations, or the like). It is noted, that an ideal amount or quantity of memory to pair with a processing component on a sled can depend upon the data center implementation as well as characteristics of the processing components, for example, the number of processing cores. The present disclosure provides memory modules having a particular amount of memory to pair with the processor components of a sled.

Reference is now made to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding thereof. It may be evident, however, that the novel embodiments can be practiced without these specific details. In other instances, known structures and devices are shown in block diagram form in order to facilitate a description thereof. The intention is to provide a thorough description such that all modifications, equivalents, and alternatives within the scope of the claims are sufficiently described.

Additionally, reference may be made to variables, such as, “a”, “b”, “c”, which are used to denote components where more than one component may be implemented. It is important to note, that there need not necessarily be multiple components and further, where multiple components are implemented, they need not be identical. Instead, use of variables to reference components in the figures is done for convenience and clarity of presentation.

FIG. 1 illustrates a conceptual overview of a data center 100 that may generally be representative of a data center or other type of computing network in/for which one or more techniques described herein may be implemented according to various embodiments. As shown in this figure, data center 100 may generally contain a plurality of racks, each of which may house computing equipment comprising a respective set of physical resources. In the particular non-limiting example depicted in this figure, data center 100 contains two racks 102A to 102B. Each of these two racks 102A to 102B may generally house a number of sleds. As shown in this figure, each of racks 102A to 102B contains four sleds 104A-1 to 104A-4 and 104B-1 to 104B-4, respectively. The depicted sleds and racks house computing equipment comprising respective sets of physical resources 105A/B. In particular, physical resources 105A-1 to 105A-4 and 105B-1 to 105B-4 are depicted. A collective set of physical resources 106 of data center 100 includes the various sets of physical resources 105 (e.g., 105A-1 to 105A-4 and 105B-1 to 105B-4) that are distributed among racks 102A to 102B.

Physical resources 106 may include resources of multiple types, such as—for example—processors, co-processors, accelerators, field-programmable gate arrays (FPGAs), memory, and storage. The embodiments are not limited to these examples. In this particular non-limiting example, physical resources 105A may thus be made up of the respective sets of physical resources housed in rack 102A, which includes physical storage resources 105A-1, physical accelerator resources 105A-2, physical memory resources 105A-3, and physical compute resources 105A-4 comprised in the sleds 104A-1 to 104A-4 of rack 102A. In some implementations, a rack may include a number of like physical resources. For example, rack 102B is depicted including physical compute resources housed in each of sleds 104B-1 to 104B-4 of rack 102B. More specifically, sleds 104B-1 to 104B-4 respectively house, physical compute resources 105B-1, physical compute resources 105B-2, physical compute resources 105B-3, and physical compute resources 105B-4.

It is noted, that embodiments are not limited to this example. Furthermore, each sled may contain a pool of each of the various types of physical resources (e.g., compute, memory, accelerator, storage). By having robotically accessible and robotically manipulatable sleds comprising disaggregated resources, each type of resource can be upgraded independently of each other and at their own optimized refresh rate.

The illustrative data center 100 differs from typical data centers in many ways. For example, in the illustrative embodiment, the circuit boards (“sleds”) on which components such as CPUs, memory, and other components are placed are designed for increased thermal performance. In particular, in the illustrative embodiment, the sleds are shallower than typical boards. In other words, the sleds are shorter from the front to the back, where cooling fans are located. This decreases the length of the path that air must to travel across the components on the board. Further, the components on the sled are spaced further apart than in typical circuit boards, and the components are arranged to reduce or eliminate shadowing (i.e., one component in the air flow path of another component). In the illustrative embodiment, processing components such as the processors are located on a top side of a sled while memory (e.g., unitary memory modules depicted herein (refer to FIGS. 6-7 and FIGS. 8A-8C)) is located on a bottom side of the sled. As a result of the enhanced airflow provided by this design, at least some components may operate at higher frequencies and power levels than in typical systems, thereby increasing performance. Furthermore, the sleds are configured to blindly mate with power and data communication cables in each rack 102A to 102B, enhancing their ability to be quickly removed, upgraded, reinstalled, and/or replaced. Similarly, individual components located on the sleds, such as processors, accelerators, memory, and data storage drives, are configured to be easily upgraded due to their increased spacing from each other. In the illustrative embodiment, the components additionally include hardware attestation features to prove their authenticity.

Furthermore, in the illustrative embodiment, the data center 100 utilizes a single network architecture (“fabric”) that supports multiple other network architectures including Ethernet and Omni-Path. The sleds, in the illustrative embodiment, are coupled to switches via optical fibers, which provide higher bandwidth and lower latency than typical twister pair cabling (e.g., Category 5, Category 5e, Category 6, etc.). Due to the high bandwidth, low latency interconnections and network architecture, the data center 100 may, in use, pool resources, such as memory, accelerators (e.g., graphics accelerators, FPGAs, ASICs, etc.), and data storage drives that are physically disaggregated, and provide them to compute resources (e.g., processors) on an as needed basis, enabling the compute resources to access the pooled resources as if they were local.

More specifically, data center 100 may feature optical fabric 112. Optical fabric 112 may generally comprise a combination of optical signaling media (such as optical cabling) and optical switching infrastructure via which any particular sled in data center 100 can send signals to (and receive signals from) each of the other sleds in data center 100. The signaling connectivity that optical fabric 112 provides to any given sled may include connectivity both to other sleds in a same rack and sleds in other racks. In the particular non-limiting example depicted in this figure, data center 100 comprises two racks (e.g., rack 102A to 102B) each including four sleds (e.g., 104A-1 to 104A-4 and 104B-1 to 104B-4, respectively). Thus, in this example, data center 100 comprises a total of eight sleds. Via optical fabric 112, each such sled may possess signaling connectivity with each of the seven other sleds in data center 100. For example, via optical fabric 112, sled 104A-1 in rack 102A may possess signaling connectivity with sled 104A-2, 104A-3 and 104A-4 in rack 102A, as well as the four other sleds 104B-1, 104B-2, 104B-3, and 104B-4 that are distributed among the other rack 102B of data center 100. The embodiments are not limited to this example.

In various embodiments, dual-mode optical switches (refer to FIG. 5 and FIG. 9) may be capable of receiving both Ethernet protocol communications carrying Internet Protocol (IP packets) and communications according to a second, high-performance computing (HPC) link-layer protocol (e.g., Intel's Omni-Path Architecture's, Infiniband) via optical signaling media of optical fabric 112. Thus, as depicted, with respect to any particular pair of sleds in data center 100, signaling connectivity via the optical fabric may provide support for link-layer connectivity via both Ethernet links and HPC links. Thus, both Ethernet and HPC communications can be supported by a single high-bandwidth, low-latency switch fabric. The embodiments are not limited to this example. However, it is worthy to note, that the dual-mode optical switches provide for separate fault domains within a single sled. As such, information can be written across fault domains at the sled level, as opposed to the rack level to provide data loss, corruption, or failure mitigation at the sled level.

The racks 102A and 102B of the data center 100 may include physical design features that facilitate the automation of a variety of types of maintenance tasks. For example, data center 100 may be implemented using racks that are designed to be robotically-accessed, and to accept and house robotically-manipulatable resource sleds. Furthermore, in the illustrative embodiment, the racks 102A and 102B include integrated power sources that receive a greater voltage than is typical for power sources. In particular examples, each of the sleds can include an associated power supply. The increased voltage enables the power sources to provide additional power to the components on each sled, enabling the components to operate at higher than typical frequencies.

As noted, the present disclosure provides sleds housing physical compute resources, such as, processor components and memory. Furthermore, the present disclosure provides a unitary memory module having a quantity or amount of memory capacity suitable to the data center in which the sled is implemented. Examples of such sleds are provides with respect to FIGS. 5-7, FIGS. 8A-8C and FIG. 9. However, a number of example racks arranged to house such sleds is given first, with respect to FIGS. 2-4. It is noted, that the term “unitary” and particularly, “unitary module” or “unitary memory module” is not meant to be limiting but is instead, relied on to reference memory module packages as described and depicted herein.

FIG. 2 illustrates a general overview of a rack architecture 200 that may be representative of an architecture of any particular one of the racks depicted in FIG. 1, according to some embodiments. As reflected in this figure, rack architecture 200 may generally feature a plurality of sled spaces into which sleds may be inserted, each of which may be robotically-accessible via a rack access region 201. In the particular non-limiting example depicted in this figure, rack architecture 200 features five sled spaces 203-1 to 203-5. Sled spaces 203-1 to 203-5 feature respective multi-purpose connector modules (MPCMs) 216-1 to 216-5. These MPCMs may be arranged to receive a corresponding MPCM of a sled (e.g., refer to FIG. 5) to mechanically, optically, and/or electrically couple the sleds to rack architecture 200, and particularly to an optical fabric of a data center to associated power sources for each sled space 203-1 to 203-5.

FIG. 3 illustrates an example of a rack architecture 300 that may be representative of a rack architecture that may be implemented in order to provide support for sleds featuring expansion capabilities (e.g., refer to FIG. 9). In the particular non-limiting example depicted in this figure, rack architecture 300 includes seven sled spaces 303-1 to 303-7, which feature respective MPCMs 316-1 to 316-7. Sled spaces 303-1 to 303-7 include respective primary regions 303-1A to 303-7A and respective expansion regions 303-1B to 303-7B. With respect to each such sled space, when the corresponding MPCM is coupled with a counterpart MPCM of an inserted sled, the primary region may generally constitute a region of the sled space that physically accommodates the inserted sled. The expansion region may generally constitute a region of the sled space that can physically accommodate an expansion module (e.g., housing additional and/or supplemental physical resources to couple with physical resources of the main sled), in the event that the inserted sled is configured with such a module.

FIG. 4 illustrates an example of a rack 402 that may be representative of a rack implemented according to rack architecture 300 of FIG. 3 according to some embodiments. In the particular non-limiting example depicted in FIG. 4, rack 402 features seven sled spaces 403-1 to 403-7, which include respective primary regions 403-1A to 403-7A and respective expansion regions 403-1B to 403-7B. In various embodiments, temperature control in rack 402 may be implemented using an air cooling system. For example, as reflected in this figure, rack 402 may feature a plurality of fans 419 that are generally arranged to provide air cooling within the various sled spaces 403-1 to 403-7. In some embodiments, the height of the sled space is greater than the conventional “1U” server height. In such embodiments, fans 419 may generally comprise relatively slow, large diameter cooling fans as compared to fans used in conventional rack configurations. Running larger diameter cooling fans at lower speeds may increase fan lifetime relative to smaller diameter cooling fans running at higher speeds while still providing the same amount of cooling. The sleds are physically shallower than conventional rack dimensions. Further, components are arranged on each sled to reduce thermal shadowing (i.e., not arranged serially in the direction of air flow). As a result, the wider, shallower sleds allow for an increase in device performance because the devices can be operated at a higher thermal envelope (e.g., 250 W) due to improved cooling (i.e., no thermal shadowing, more space between devices, more room for larger heat sinks, etc.).

MPCMs 416-1 to 416-7 may be configured to provide inserted sleds with access to power sourced by respective power modules 420-1 to 420-7, each of which may draw power from an external power source 421. In various embodiments, external power source 421 may deliver alternating current (AC) power to rack 402, and power modules 420-1 to 420-7 may be configured to convert such AC power to direct current (DC) power to be sourced to inserted sleds. In some embodiments, for example, power modules 420-1 to 420-7 may be configured to convert 277-volt AC power into 12-volt DC power for provision to inserted sleds via respective MPCMs 416-1 to 416-7. The embodiments are not limited to this example.

MPCMs 416-1 to 416-7 may also be arranged to provide inserted sleds with optical signaling connectivity to an optical fabric, which may be the same as—or similar to—optical fabric 112 of FIG. 1. In various embodiments, optical connectors contained in MPCMs 416-1 to 416-7 may be designed to couple with counterpart optical connectors contained in MPCMs of inserted sleds to provide such sleds with optical signaling connectivity to optical fabric 412 via respective lengths of optical cabling 422-1 to 422-7. In some embodiments, each such length of optical cabling may extend from its corresponding MPCM to an optical interconnect loom 423 that is external to the sled spaces of rack 402. In various embodiments, optical interconnect loom 423 may be arranged to pass through a support post or other type of load-bearing element of rack 402. The embodiments are not limited in this context. Because inserted sleds connect to an optical switching infrastructure via MPCMs, the resources typically spent in manually configuring the rack cabling to accommodate a newly inserted sled can be saved.

FIG. 5 illustrates an example of a sled 504 that may be representative of a sled designed for use in conjunction with a rack according to some embodiments (e.g., racks according to rack architectures 200 or 300 or rack 402). Sled 504 may feature an MPCM 516 that comprises an optical connector 516A and a power connector 516B, and that is designed to couple with a counterpart MPCM of a sled space in conjunction with insertion of MPCM 516 into that sled space. Coupling MPCM 516 with such a counterpart MPCM may cause power connector 516B to couple with a power connector comprised in the counterpart MPCM. This may generally enable physical compute resources 505 of sled 504 to source power from an external source, via power connector 516B and power transmission media 524 that conductively couples power connector 516B to physical compute resources 505.

Physical compute resources 505 can generally include any number of processor component and associated memory. For example, physical compute resources 505 includes processor components 533-1 and 533-2 and memory 535-1 and 535-2. Processor component 533-1 is operably coupled to memory 535-1 via electrical signaling media 528 while processor component 533-2 is operably coupled to memory 535-2 via electrical signaling media 528.

In general, processor components 533-1 can be any of a variety of processors, such as, central processing units (CPUs), graphics processing units (GPUs), field-programmable gate arrays (FPGAs) or the like. In this illustrative example, processor components 533-1 and 533-2 can be central processing units comprising a number of processing cores. For example, each of processing components 533-1 and 533-2 can have any number of cores, even a different number of cores. As a specific example, each of processing component 633-1 and 533-2 can have 2 cores, 4 cores, 8 cores, 12 cores, 24 cores, 32 cores, or the like. In this illustrative example, processor components 533-1 and 533-2 are depicted including 4 cores each. Specifically, processor component 533-1 is depicted including 4 cores 580-1 while processor component 533-2 is depicted including 4 cores 580-2. Examples are however, not limited in this context. Furthermore, processing components 533-1 and 533-2 can be an x86 (e.g., 32 bit, 64, bit, or the like) based processor manufactured at any of a variety of device fabrication nodes, such as, for example, 7 nanometer (nm) node, 10 nm node, 14 nm node, 22 nm, 32 nm, 45 nm node, or the like. Furthermore, the processing components can be packed in any of a variety of package types having various pin counts. Examples are not limited in this context.

As will be described in greater detail below, memory 535-1 and 535-2 can be embodied in a ball grid array (BGA) package and referred to as a “unitary module” or a “unitary memory module.” Furthermore, each of the unitary memory modules 535-1 and 535-2 can include a controller (e.g., memory controller, or the like) and a memory. For example, unitary memory module 535-1 can include controller 590-1 and memory 592-1 while unitary memory module 535-2 can include controller 590-2 and memory 592-2. In general, memory 535-1 and memory 535-2 (or more particularly, memory 592-1 and 592-2) can be any of a variety of types of memory, including volatile memory, non-volatile memory, etc.

In some examples, sled 502 can comprise two levels of memory (sometimes referred to as ‘2LM’). A first level of the 2LM architecture can comprise smaller faster memory while a second level of memory can comprise larger and slower memory, relative to the first level. In some cases, the first level of memory can be referred to as near memory while the second level of memory can be referred to as far memory. With some examples, the unitary memory modules 535-1 and 535-2 can be implemented as near memory for corresponding processor components 533-1 and 533-2.

For example, unitary memory modules 535-1 and 535-2 (and particularly memory 592-1 and 592-2) can be implemented from random-access memory (RAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-data rate SDRAM, NAND memory, NOR memory, three-dimensional (3D) cross-point memory, ferroelectric memory, silicon-oxide-nitride-oxide-silicon (SONOS) memory, polymer memory such as ferroelectric polymer memory, ferroelectric transistor random access memory (FeTRAM or FeRAM), nanowire, phase-change RAM (PRAM), resistive RAM (RRAM), magnetoresistive RAM (MRAM), spin transfer torque MRAM (STT-MRAM) memory, non-volatile static RAM (nvSRAM), conductive-bridging RAM (CBRAM), nano-RAM (NRAM), floating junction gate RAM (FJG RAM), or the like.

Unitary memory modules 535-1 and 535-2 can include a quantity of memory (e.g., memory 592-1 and 592-2, respectively) based on processor component 533-1 and 533-2 to which unitary memory modules are coupled. For example, unitary memory modules 535-1 and 535-2 can include between 2 and 4 Gigabytes (GB) of memory for each core of processor component 533-1 or 533-2 to which unitary memory module 535-1 and 535-2 are attached. As a specific example, each of processor components 533-1 and 533-2 can include 32 cores while each unitary memory module includes 96 GB of memory 533-1 and 533-2, which equates to 3 GB of memory per core.

The present disclosure can provide a sled having unitary memory modules (e.g., unitary memory modules 535-1 and 535-2, or the like) arranged and configured to be removed in an autonomous process, such as, for example, by a robot operating in a data center. Examples are not limited in this context.

Sled 504 may also include dual-mode optical network interface circuitry 526. Dual-mode optical network interface circuitry 526 may generally comprise circuitry that is capable of communicating over optical signaling media according to each of multiple link-layer protocols supported by an optical fabric (e.g., optical fabric 112 of FIG. 1, optical fabric 414 of FIG. 4, or the like). In some embodiments, dual-mode optical network interface circuitry 526 may be capable both of Ethernet protocol communications and of communications according to a second, high-performance protocol. In various embodiments, dual-mode optical network interface circuitry 526 may include one or more optical transceiver modules 527, each of which may be capable of transmitting and receiving optical signals over each of one or more optical channels. The embodiments are not limited in this context.

Coupling MPCM 516 with a counterpart MPCM of a sled space in a given rack may cause optical connector 516A to couple with an optical connector comprised in the counterpart MPCM. This may generally establish optical connectivity between optical cabling of the sled and dual-mode optical network interface circuitry 526, via each of a set of optical channels 525. With some examples, optical channels 525 comprise 4 optical fiber channels. With some examples, each of the optical channels can provide between 20 and 220 Gigabytes per second (GB/s) bandwidth. With a specific example, each of the optical channels can provide 50 GB/s bandwidth. As another specific example, each of the optical channels can provide 200 GB/s bandwidth. Dual-mode optical network interface circuitry 526 may communicate with the physical resources 505 of sled 504 via electrical signaling media 528. In addition to the dimensions of the sleds and arrangement of components on the sleds to provide improved cooling and enable operation at a relatively higher thermal envelope (e.g., 250 W), as described above with reference to FIG. 4, in some embodiments, a sled may include one or more additional features to facilitate air cooling, such as a heat pipe and/or heat sinks arranged to dissipate heat generated by physical resources 505. It is worthy of note that although the example sled 504 depicted in FIG. 5 does not feature an expansion connector, any given sled that features the design elements of sled 504 may also feature an expansion connector according to some embodiments. The embodiments are not limited in this context.

FIG. 6 depicts a perspective view of an example sled 604. It is noted, that the example sled 604 is not depicted to scale and particularly, features of sled 604 are depicted in exaggerated form to facilitate understanding. The example sled 604 includes a substrate 640, processor components 633-1 and 633-2, and unitary memory modules 635-1 and 635-2. Processor components 633-1 and 633-2 can be any of a variety of processor components, and can be like—the processor components 533-1 and 533-2 depicted and described with respect to FIG. 5. Furthermore, sled 604 can include processor component heat sinks (refer to FIG. 7) thermally coupled to processor components 633-1 and 633-2 to dissipate thermal energy generated by processor component 633-1 and 633-2 during operation.

Unitary memory modules 635-1 and 635-2 can be memory (e.g., DRAM, or the like) packaged in unitary modules and coupled to respective ones of processor components 633-1 and 633-2. In general, unitary memory modules 635-1 and 635-2 can be packaged into a BGA package suitable to couple to a socket disposed on substrate 640 (refer to FIG. 8A-8C). Sled 604 can include unitary module heat sinks (refer to FIG. 7) thermally coupled to unitary memory modules 635-1 and 635-2 to dissipate thermal energy generated by unitary memory modules 635-1 and 635-2 during operation.

In general, processor components 633-1 and 633-2 (and associated heat sinks) can be disposed on a first side (e.g., upper surface in this example) of substrate 640. Furthermore, unitary memory modules 635-1 and 635-2 can be disposed on a second side (e.g., lower surface in this example) of substrate 640. It is noted, that the first side and second side, or first surface of substrate and second surface of substrate 640, to which processor components 633-1 and 633-2 and unitary memory modules 635-1 and 635-2 are respectively coupled can be opposite from each other. Said differently, computing resources (e.g., processor components, or the like) can be disposed on the upper surface of the sled 604 while memory for processor components (e.g., unitary memory modules, or the like) can be disposed on the lower surface of sled 604. The sled 604 can further comprise (not shown) circuit boards and/or connective components to provide connectivity between the processor components and memory, as well as other interconnects of the sled 604 (e.g., optical interconnects, or the like).

FIG. 7 depicts a perspective view of an example sled 704. It is noted, that the example sled 704 is not depicted to scale. The example sled 704 includes a substrate 704, processor components 733-1 and 733-2 (obscured in this view), unitary memory module 735-1, unitary memory module 735-2 (obscured in this view), processor component heat sinks 737-1 and 737-2, unitary module heat sink 739-1 (not shown) and unitary module heat sink 739-2.

Processor components 733-1 and 733-2 can be any of a variety of processor components, and can, be like, the processor components 533-1 and 533-2 depicted and described with respect to FIG. 5. Furthermore, sled 704 can include processor component heat sinks 737-1 and 737-2 thermally coupled to respective processor components 733-1 and 733-2 to dissipate thermal energy generated by processor components 733-1 and 733-2 during operation. It is noted, that processor components 733-1 and 733-2 are obscured from view by the processor component heat sinks 737-1 and 737-2.

Unitary memory modules 735-1 and 735-2 can be memory (e.g., DRAM, or the like) packaged in unitary modules and coupled to respective ones of processor components 733-1 and 733-2. In general, unitary memory modules 735-1 and 735-2 can be packaged into a BGA package suitable to couple to a socket disposed on substrate 740 (refer to FIG. 8A-8C). For example, unitary module BGA package 750-1 is depicted comprising a memory array 752-1 on one side and BGA contacts (obscured in this view) on a side facing the substrate 740. As described above, unitary memory modules can include a quantity of memory based in part on the number of cores of a processor component to which the unitary memory module is coupled. As such, in this illustrative example, memory array 752-1 could include a quantity of memory (e.g., GBs of DRAM, or the like) based in part on the number of cores of processor component 733-1.

Sled 704 can include unitary module heat sinks 739-1 (not shown) and 739-2 thermally coupled to unitary memory modules 735-1 and 735-2, respectively. Unitary module heat sinks 739-1 and 739-2 can dissipate thermal energy generated by unitary memory modules 735-1 and 735-2 during operation.

In general, heat sinks (e.g., 737-1, 737-2, 739-1, 739-2, or the like) can be mechanically coupled to substrate 740 and thermally coupled to an active component (e.g., processor component 733-1, processor component 733-2, unitary memory module 735-1, unitary memory module 735-2, or the like) via any of a variety of methods, such as, for example, screws, hold-downs, springs, frames, buttons, thermal paste, surface contact, or the like.

In general, processor components 733-1 and 733-2 and associated heat sinks 737-1 and 737-2 can be disposed on a first side (e.g., upper surface in this example) of substrate 740. Furthermore, unitary memory modules 735-1 and 735-2 and associated heat sinks 739-1 and 739-2 can be disposed on a second side (e.g., lower surface in this example) of substrate 740. It is noted, that the first side and second side, or first surface of substrate and second surface of substrate to which processor components 733-1 and 733-2 and unitary memory modules 735-1 and 735-2 are respectively coupled can be opposite from each other. Said differently, computing resources (e.g., processor components, or the like) can be disposed on the upper surface of the sled 704 while memory for processor components (e.g., unitary memory modules, or the like) can be disposed on the lower surface of sled 704. The sled 704 can further comprise (not shown) circuit boards and/or connective components to provide connectivity between the processor components and memory, as well as other interconnects of the sled 704 (e.g., optical interconnects, or the like).

FIGS. 8A-8C depict a perspective view of an example sled 804 and removal (or installation) of a unitary memory module. It is noted, that the example sled 804 is not depicted to scale and does not depict all elements that can be implemented on such an example sled. In particular, this illustrative example depicts sled 804 (or a portion of sled 804) including substrate 840 and features a unitary memory module. Turning more specifically to FIG. 8A, substrate 840 is depicted with unitary module heat sink 839 coupled to substrate 840. In particular, unitary module heat sink 839 is coupled to substrate 840 via heat sink fasteners 862. In some examples, heat sink fasteners 862 can be arranged and/or configured to provide automated removal and installation of heat sink 839 to substrate 840. For example, heat sink 839 and fasteners 862 can be arranged and configured to be removed by an autonomous apparatus, such as a robot.

Turning more specifically to FIG. 8B, substrate 840 is depicted with unitary module heat sink 839 removed such that unitary module package 850 is illustrated. As depicted, unitary module package 850 is coupled to substrate 840. Unitary module package is coupled to substrate 840 via a hinge 870 and a frame 872. In some examples, frame 872 can be arranged and/or configured to provide automated removal and installation of unitary module package 850 to substrate 840. For example, frame 872 can be manipulated about hinge 870 by an autonomous apparatus, such as a robot.

Turning more specifically to FIG. 8C, substrate 840 is depicted with unitary module package 850 removed such that BGA contacts 842 are illustrated. As depicted, BGA contacts 842 are arranged on substrate 840 to provide electrical coupling between unitary module package 850 and a corresponding processor component of sled 804. It is noted, that BGA contacts 842 are depicted in an arrangement (e.g., row, column, etc.) and at a quantity to facilitate understanding. However, in practice, the BGA array 842 can have any shape and number of individual contacts suitable to operably couple unitary modules to processor components. Examples are not limited in this context.

As noted above, with some examples, a sled can be arranged to accept an expansion sled. FIG. 9 illustrates an example of a sled 904 that may be representative of a sled of such a type. As shown in this figure, sled 904 may comprise a set of physical resources 905, as well as an MPCM 916 designed to couple with a counterpart MPCM when sled 904 is inserted into a sled space such as any of sled spaces 303-1 to 303-7 of FIG. 3. Sled 904 can also feature dual-mode optical network interface circuitry 926 to couple components of sled 904 to optical fabric of a data center.

Sled 904 may also feature an expansion connector 917. Expansion connector 917 may generally comprise a socket, slot, or other type of connection element that is capable of accepting one or more types of expansion modules, such as an expansion sled 918. By coupling with a counterpart connector on expansion sled 918, expansion connector 917 may provide physical resources 905 with access to supplemental physical resources 905B residing on expansion sled 918.

For example, physical resources 905 can comprise physical compute resources, such as, processor component(s) 933 and unitary memory module(s) 935. Additional processor components (e.g., co-processor, accelerators, GPUS processors, or the like) or memory (e.g., far memory, or the like) to be included in physical resources can be provided via supplemental physical resources 905B on expansion sled 918.

FIG. 10 illustrates an example of a data center 1000 that may generally be representative of one in/for which one or more techniques described herein may be implemented according to various embodiments. As reflected in this figure, a physical infrastructure management framework 1050A may be implemented to facilitate management of a physical infrastructure 1000A of data center 1000. In various embodiments, one function of physical infrastructure management framework 1050A may be to manage automated maintenance functions within data center 1000, such as the use of robotic maintenance equipment to service computing equipment within physical infrastructure 1000A. In some embodiments, physical infrastructure 1000A may feature an advanced telemetry system that performs telemetry reporting that is sufficiently robust to support remote automated management of physical infrastructure 1000A. In various embodiments, telemetry information provided by such an advanced telemetry system may support features such as failure prediction/prevention capabilities and capacity planning capabilities. In some embodiments, physical infrastructure management framework 1050A may also be configured to manage authentication of physical infrastructure components using hardware attestation techniques. For example, robots may verify the authenticity of components before installation by analyzing information collected from a radio frequency identification (RFID) tag associated with each component to be installed. The embodiments are not limited in this context.

As shown in this figure, the physical infrastructure 1000A of data center 1000 may comprise an optical fabric 1012, which may include a dual-mode optical switching infrastructure 1014. Optical fabric 1012 and dual-mode optical switching infrastructure 1014 may be the same as—or similar to—optical fabric 102 of FIG. 1 or 412 of FIG. 4, and may provide high-bandwidth, low-latency, multi-protocol connectivity among sleds of data center 1000. As discussed above, with reference to FIG. 1, in various embodiments, the availability of such connectivity may make it feasible to disaggregate and dynamically pool resources such as accelerators, memory, and storage. In some embodiments, for example, one or more pooled accelerator sleds 1030 may be included among the physical infrastructure 1000A of data center 1000, each of which may comprise a pool of accelerator resources—such as co-processors and/or FPGAs, for example—that is available globally accessible to other sleds via optical fabric 1012 and dual-mode optical switching infrastructure 1014.

In another example, in various embodiments, one or more pooled storage sleds 1032 may be included among the physical infrastructure 1000A of data center 1000, each of which may comprise a pool of storage resources that is available globally accessible to other sleds via optical fabric 1012 and dual-mode optical switching infrastructure 1014. In some embodiments, such pooled storage sleds 1032 may comprise pools of solid-state storage devices such as solid-state drives (SSDs). In various embodiments, one or more high-performance processing sleds 1034 may be included among the physical infrastructure 1000A of data center 1000. In some embodiments, high-performance processing sleds 1034 may comprise pools of high-performance processors, as well as cooling features that enhance air cooling to yield a higher thermal envelope of up to 250 W or more. In various embodiments, any given high-performance processing sled 1034 may feature an expansion connector 1017 that can accept a far memory expansion sled, such that the far memory that is locally available to that high-performance processing sled 1034 is disaggregated from the processors and memory comprised on that sled. In some embodiments, such a high-performance processing sled 1034 may be configured with far memory using an expansion sled that comprises low-latency SSD storage. The optical infrastructure allows for compute resources on one sled to utilize remote accelerator/FPGA, memory, and/or SSD resources that are disaggregated on a sled located on the same rack or any other rack in the data center. The remote resources can be located one switch jump away or two-switch jumps away in the spine-leaf network architecture. The embodiments are not limited in this context.

In various embodiments, one or more layers of abstraction may be applied to the physical resources of physical infrastructure 1000A in order to define a virtual infrastructure, such as a software-defined infrastructure 1000B. In some embodiments, virtual computing resources 1036 of software-defined infrastructure 1000B may be allocated to support the provision of cloud services 1040. In various embodiments, particular sets of virtual computing resources 1036 may be grouped for provision to cloud services 1040 in the form of SDI services 1038. Examples of cloud services 1040 may include—without limitation—software as a service (SaaS) services 1042, platform as a service (PaaS) services 1044, and infrastructure as a service (IaaS) services 1046.

In some embodiments, management of software-defined infrastructure 1000B may be conducted using a virtual infrastructure management framework 1050B. In various embodiments, virtual infrastructure management framework 1050B may be designed to implement workload fingerprinting techniques and/or machine-learning techniques in conjunction with managing allocation of virtual computing resources 1036 and/or SDI services 1038 to cloud services 1040. In some embodiments, virtual infrastructure management framework 1050B may use/consult telemetry data in conjunction with performing such resource allocation. In various embodiments, an application/service management framework 1050C may be implemented in order to provide QoS management capabilities for cloud services 1040. The embodiments are not limited in this context.

One or more aspects of at least one example may be implemented by representative instructions stored on at least one machine-readable medium which represents various logic within the processor, which when read by a machine, computing device or system causes the machine, computing device or system to fabricate logic to perform the techniques described herein. Such representations, known as “IP cores” may be stored on a tangible, machine readable medium and supplied to various customers or manufacturing facilities to load into the fabrication machines that actually make the logic or processor.

Various examples may be implemented using hardware elements, software elements, or a combination of both. In some examples, hardware elements may include devices, components, processors, microprocessors, circuits, circuit elements (e.g., transistors, resistors, capacitors, inductors, and so forth), integrated circuits, application specific integrated circuits (ASICs), programmable logic devices (PLDs), digital signal processors (DSPs), field programmable gate array (FPGA), memory units, logic gates, registers, semiconductor device, chips, microchips, chip sets, and so forth. In some examples, software elements may include software components, programs, applications, computer programs, application programs, system programs, machine programs, operating system software, middleware, firmware, software modules, routines, subroutines, functions, methods, procedures, software interfaces, application program interfaces (API), instruction sets, computing code, computer code, code segments, computer code segments, words, values, symbols, or any combination thereof. Determining whether an example is implemented using hardware elements and/or software elements may vary in accordance with any number of factors, such as desired computational rate, power levels, heat tolerances, processing cycle budget, input data rates, output data rates, memory resources, data bus speeds and other design or performance constraints, as desired for a given implementation.

Some examples may include an article of manufacture or at least one computer-readable medium. A computer-readable medium may include a non-transitory storage medium to store logic. In some examples, the non-transitory storage medium may include one or more types of computer-readable storage media capable of storing electronic data, including volatile memory or non-volatile memory, removable or non-removable memory, erasable or non-erasable memory, writeable or re-writeable memory, and so forth. In some examples, the logic may include various software elements, such as software components, programs, applications, computer programs, application programs, system programs, machine programs, operating system software, middleware, firmware, software modules, routines, subroutines, functions, methods, procedures, software interfaces, API, instruction sets, computing code, computer code, code segments, computer code segments, words, values, symbols, or any combination thereof.

According to some examples, a computer-readable medium may include a non-transitory storage medium to store or maintain instructions that when executed by a machine, computing device or system, cause the machine, computing device or system to perform methods and/or operations in accordance with the described examples. The instructions may include any suitable type of code, such as source code, compiled code, interpreted code, executable code, static code, dynamic code, and the like. The instructions may be implemented according to a predefined computer language, manner or syntax, for instructing a machine, computing device or system to perform a certain function. The instructions may be implemented using any suitable high-level, low-level, object-oriented, visual, compiled and/or interpreted programming language.

Some examples may be described using the expression “in one example” or “an example” along with their derivatives. These terms mean that a particular feature, structure, or characteristic described in connection with the example is included in at least one example. The appearances of the phrase “in one example” in various places in the specification are not necessarily all referring to the same example.

Some examples may be described using the expression “coupled” and “connected” along with their derivatives. These terms are not necessarily intended as synonyms for each other. For example, descriptions using the terms “connected” and/or “coupled” may indicate that two or more elements are in direct physical or electrical contact with each other. The term “coupled,” however, may also mean that two or more elements are not in direct contact with each other, but yet still co-operate or interact with each other.

It is emphasized that the Abstract of the Disclosure is provided to comply with 37 C.F.R. Section 1.72(b), requiring an abstract that will allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in a single example for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed examples require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed example. Thus the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate example. In the appended claims, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein,” respectively. Moreover, the terms “first,” “second,” “third,” and so forth, are used merely as labels, and are not intended to impose numerical requirements on their objects.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

The present disclosure can be implemented in any of a variety of embodiments, such as, for example, the following non-exhaustive listing of example embodiments.

Example 1

An apparatus for a sled to house physical compute resources of a data center, the apparatus comprising: a substrate; a first socket to receive a processor component, the first socket disposed on a first surface of the substrate; and a first memory socket to receive a memory module, the first memory socket disposed on a second surface of the substrate different than the first surface of the substrate, the first memory socket to couple the memory module to the processor component.

Example 2

The apparatus of example 1, wherein the first memory socket is configured to receive a unitary memory module.

Example 3

The apparatus of example 2, comprising the processor component and the unitary memory module.

Example 4

The apparatus of example 3, the unitary memory module comprising a quantity of memory based in part on a number of cores of the processor component.

Example 5

The apparatus of example 3, comprising a processor component heat sink mechanically coupled to the substrate and thermally coupled to the processor component.

Example 6

The apparatus of example 5, comprising a unitary memory module heat sink mechanically coupled to the substrate and thermally coupled to the unitary memory module.

Example 7

The apparatus of example 6, comprising the unitary memory module heat sink removably mechanically coupled to the substrate.

Example 8

The apparatus of example 7, comprising a hinge coupled to the substrate and a frame coupled to the hinge, the frame and hinge to removably mechanically couple the unitary memory module to the substrate.

Example 9

The apparatus of example 8, the frame and hinge to removably mechanically couple the unitary memory module and the unitary memory module heat sink to the substrate.

Example 10

The apparatus of example 2, comprising: a second socket to receive a processor component, the second socket disposed on the first surface of the substrate; and a second memory socket to receive a unitary memory module, the second memory socket disposed on the second surface of the substrate, the second memory socket to couple the unitary memory module of the second memory socket to the processor component of the second socket.

Example 11

The apparatus of any one of examples 1 to 10, the first memory socket comprising a ball grid array (BGA) socket.

Example 12

The apparatus of any one of examples 1 to 10, the first surface and the second surface opposite from each other.

Example 13

The apparatus of example 3, the processor component comprising between 2 and 32 cores.

Example 14

The apparatus of example 13, the quantity of memory comprising between 1 and 4 gigabytes of memory per core.

Example 15

The apparatus of any one of examples 2 to 10, the unitary memory module comprising dynamic random access memory (DRAM) or three-dimensional (3D) cross-point memory.

Example 16

A system for a data center comprising: a rack comprising a plurality of sled spaces; and at least one sled coupled to the rack via a one of the plurality of sled spaces, the sled comprising: a substrate; a first socket to receive a processor component, the first socket disposed on a first surface of the substrate; and a first memory socket to receive a memory module, the first memory socket disposed on a second surface of the substrate different than the first surface of the substrate, the first memory socket to couple the memory module to the processor component.

Example 17

The system of example 16, wherein the first memory socket is configured to receive a unitary memory module.

Example 18

The system of example 17, the sled comprising the processor component and the unitary memory module.

Example 19

The system of example 18, the unitary memory module comprising a quantity of memory based in part on a number of cores of the processor component.

Example 20

The system of example 18, the sled comprising a processor component heat sink mechanically coupled to the substrate and thermally coupled to the processor component.

Example 21

The system of example 20, the sled comprising a unitary memory module heat sink mechanically coupled to the substrate and thermally coupled to the unitary memory module.

Example 22

The system of example 21, the sled comprising the unitary memory module heat sink removably mechanically coupled to the substrate.

Example 23

The system of example 22, the sled comprising a hinge coupled to the substrate and a frame coupled to the hinge, the frame and hinge to removably mechanically couple the unitary memory module to the substrate.

Example 24

The system of example 23, the frame and hinge to removably mechanically couple the unitary memory module and the unitary memory module heat sink to the substrate.

Example 25

The system of example 17, the sled comprising: a second socket to receive a processor component, the second socket disposed on the first surface of the substrate; and a second memory socket to receive a unitary memory module, the second memory socket disposed on the second surface of the substrate, the second memory socket to couple the unitary memory module of the second memory socket to the processor component of the second socket.

Example 26

The system of any one of examples 17 to 25, the first memory socket comprising a ball grid array (BGA) socket.

Example 27

The system of any one of examples 17 to 25, the first surface and the second surface opposite from each other.

Example 28

The system of example 18, the processor component comprising between 2 and 32 cores.

Example 29

The system of example 28, the quantity of memory comprising between 1 and 4 gigabytes of memory per core.

Example 30

The system of any one of examples 17 to 25, the memory module comprising dynamic random access memory (DRAM) or three-dimensional (3D) cross-point memory.

Example 31

An apparatus for a physical resource sled in a data center, comprising: a substrate mountable within a sled space of a rack of a data center; a plurality of sockets coupled to the substrate, each of the plurality of sockets to receive a processor component; and a memory module for each of the plurality of sockets, the memory module communicatively coupled to a respective socket to couple the memory module to a processor component received by the socket, each of the memory modules comprising: a quantity of memory based in part on a number of cores of the processor component; and a memory controller to couple the quantity of memory to processor component.

Example 32

The apparatus of example 31, wherein the plurality of memory modules comprising a unitary memory module.

Example 33

The apparatus of example 32, comprising the plurality of processor components.

Example 34

The apparatus of example 33, comprising a plurality of processor component heat sinks mechanically coupled to the substrate, each of the plurality of processor component heat sinks thermally coupled to a respective one of the plurality of processor components.

Example 35

The apparatus of example 34, comprising a plurality of unitary memory module heat sinks mechanically coupled to the substrate, each of the plurality of unitary memory module heat sinks thermally coupled to a respective one of the plurality of unitary memory modules.

Example 36

The apparatus of example 35, comprising the plurality of unitary memory module heat sinks removably mechanically coupled to the substrate.

Example 37

The apparatus of example 36, comprising a hinge coupled to the substrate and a frame coupled to the hinge, the frame and hinge to removably mechanically couple the plurality of unitary memory modules to the substrate.

Example 38

The apparatus of example 37, the frame and hinge to removably mechanically couple the plurality of unitary memory modules and the plurality of unitary memory module heat sinks to the substrate.

Example 39

The apparatus of any one of examples 32 to 38, the plurality of sockets disposed on a first surface of the substrate and the plurality of unitary memory modules disposed on a second surface of the substrate.

Example 40

The apparatus of example 39, the first surface opposite from the second surface.

Example 41

The apparatus of any one of examples 32 to 38, each of the plurality of processor components comprising between 2 and 32 cores.

Example 42

The apparatus of example 41, the quantity of memory comprising between 1 and 4 gigabytes of memory per core.

Example 43

The apparatus of any one of examples 32 to 38, the memory comprising dynamic random access memory (DRAM) or three-dimensional (3D) cross-point memory.

Example 44

A method for a sled of a rack of a data center, the method comprising: receiving a processor component at a first socket, the first socket disposed on a first surface of a substrate of a sled; receiving a memory module at a first memory socket, the first memory socket disposed on a second surface of the substrate different than the first surface; and coupling, via the first socket and the first memory socket, the memory module to the processor component.

Example 45

The method of example 44, comprising receiving a unitary memory module at the first memory socket.

Example 46

The method of example 45, the unitary memory module comprising a quantity of memory based in part on a number of cores of the processor component.

Example 47

The method of example 45, the sled comprising a processor component heat sink mechanically coupled to the substrate and thermally coupled to the processor component.

Example 48

The method of example 47, the sled comprising a unitary memory module heat sink mechanically coupled to the substrate and thermally coupled to the unitary memory module.

Example 49

The method of example 48, comprising removing the unitary memory module heat sink from the substrate.

Example 50

The method of example 48, the sled comprising a hinge coupled to the substrate and a frame coupled to the hinge, the frame and hinge to removably mechanically couple the unitary memory module to the substrate.

Example 51

The method of example 50, the frame and hinge to removably mechanically couple the unitary memory module and the unitary memory module heat sink to the substrate.

Example 52

The method of example 50, comprising: receiving a processor component at a second socket, the second socket disposed on the first surface of the substrate of the sled; receiving a memory module at a second memory socket, the second memory socket disposed on the second surface of the substrate; and coupling, via the second socket and the second memory socket, the memory module to the processor component.

Example 53

The method of any one of examples 45 to 52, the first memory socket comprising a ball grid array (BGA) socket.

Example 54

The method of any one of examples 45 to 52, the first surface and the second surface opposite from each other.

Example 55

The method of example 44, the processor component comprising between 2 and 32 cores.

Example 56

The method of example 55, the quantity of memory comprising between 1 and 4 gigabytes of memory per core.

Example 57

The method of any one of examples 45 to 52, the memory module comprising dynamic random access memory (DRAM) or three-dimensional (3D) cross-point memory.

Number	Date	Country
62427268	Nov 2016	US
62376859	Aug 2016	US
62365969	Jul 2016	US

Memory Module for a Data Center Compute Sled

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

International Classifications

Abstract

Description

Claims

RELATED APPLICATIONS

Provisional Applications (3)