Versatile SMP (vSMP) Architecture


Introducing High-end Virtualization

Computing virtualization is a technique for hiding the physical characteristics of a compute resource from the OS, applications or end users interacting with that compute resource. There are two types of computing virtualization paradigms in the market today: server virtualization, where a single physical server appears to function as multiple logical (virtual) servers (can also be defined as “partitioning”); and desktop virtualization, in which the physical location of the PC desktop is separated from the user that is accessing the PC. (Such a remotely accessed PC can be located at the user’s home, office or data center, while the user is located elsewhere; it can also be defined as “remoting.”)

ScaleMP has created a new computing virtualization paradigm: high-end virtualization. With high-end virtualization, multiple physical systems appear to function as a single logical system. ScaleMP defines this virtualization paradigm as “aggregation,” considering it is basically the opposite of “partitioning.” By aggregating the physical resources, all of these resources are available to all of the applications.

ScaleMP is the leader in virtualization for aggregation. The company’s innovative Versatile SMP (vSMP) architecture aggregates multiple x86 systems into a single virtual x86 system, delivering an industry-standard, high-end symmetric multiprocessor (SMP) computer. ScaleMP leverages software to replace custom hardware and components to offer a new, revolutionary computing paradigm.

Partitioning vs. Aggregation

Back to top


The Versatile SMP (vSMP) Architecture

ScaleMP’s patented Versatile SMP (vSMP) architecture enables the creation of high-end SMP systems. It fundamentally replaces the functionality of custom and proprietary chipsets with software and commodity interconnects such as InfiniBand. And because it utilizes only a small fraction of the system’s CPUs and RAM to provide chipset-level services, system performance is never sacrificed.

vSMP Foundation is ScaleMP’s implementation of the vSMP architecture. vSMP Foundation aggregates multiple x86 servers into one larger SMP system, allowing system vendors and value-added resellers to create high-end x86 solutions using industry-standard components. This eliminates the need for lengthy and onerous custom hardware development.

Traditional multi-processor systems require the creation of a custom chipset and board to enable processor communication and shared memory. The larger the system is, the more complex the required solution is. Chipsets that support more than four processors are complicated to design, and very few implementations exist. Moreover, as the technology refresh cycle in the x86 space is 18 months or less, chipsets and boards require significant ongoing investments to keep up with the advancement of technology. This inevitably results in slower technology adaptations in the high-end market and more expensive, low-performing systems.

The vSMP architecture utilizes off-the-shelf components and does not require any custom parts. The key value of the vSMP architecture is the utilization of software to provide the chipset services that are otherwise required in creating traditional multi-processor systems. The vSMP architecture provides cache coherency, shared I/O and the system interfaces (BIOS, ACPI), which are required by the OS. It is implemented in a completely transparent manner; no additional device drivers are required and no modifications to the OS or the applications are necessary.

Back to top


Requirements

vSMP Foundation Requirements:

  • Multiple high-volume, industry-standard x86 systems or system boards with processors and memory (processor speed and amount of memory across boards does not have to be the same)
  • InfiniBand infrastructure in the form of HCAs, cables and switches
  • vSMP Foundation persistent storage devices (one per board) that are used to boot the system board into vSMP Foundation. The devices are plugged into each system board and are loaded with the appropriate vSMP Foundation product. These vSMP Foundation Devices are not required for vSMP Foundation for Cloud. The Cloud product is network-bootable.

 

Back to top


One System

vSMP Foundation 5.0 supports the aggregation of up to 128 x86 server systems to create one system. Once loaded into the memory of each of the system boards, vSMP Foundation aggregates the compute, memory and I/O capabilities of each system and presents a unified virtual system to both the OS and the applications running above the OS. vSMP Foundation uses a software-interception engine in the form of a Virtual Machine Monitor to provide a uniform execution environment. vSMP Foundation also creates the required BIOS and ACPI environment to provide the OS (and the software stack above the OS) a coherent image of a single system.

Back to top


Coherent Memory

vSMP Foundation maintains cache coherency between the individual boards using multiple advanced coherency algorithms. These complex algorithms operate concurrently on a per-block basis, based on real-time memory activity access patterns. vSMP Foundation leverages board local-memory together with best-of-breed caching algorithms to minimize the effect of interconnect latencies. This gives a vSMP Foundation system coherent memory.

Back to top


Shared I/O

vSMP Foundation aggregates I/O resources across all boards into a unified PCI hierarchy and presents them as a common pool of I/O resources to the OS and the application. The OS is able to utilize all the system storage and networking controllers towards providing high-I/O system capabilities.

Back to top


Flexible Versatile Architecture

vSMP Foundation Advanced Platform software implements the vSMP architecture, enabling the creation of a single virtual system by aggregating up to 128 industry-standard x86 systems. vSMP Foundation creates virtual systems ranging from four to 1,024 processors (32,768 CPUs) and up to 256 TB of shared memory.

vSMP Foundation aggregates system boards with different processor speeds, varied memory amounts or dissimilar I/O devices. This is a unique capability among x86 shared-memory systems.

For applications that are compute-intensive, it is possible to create homogeneous systems delivering up to 300 TFLOPS. For applications that are memory-intensive and not compute-intensive, an imbalanced configuration that mixes high-speed and low-speed processors can be architected. With such an imbalanced configuration, vSMP Foundation will aggregate only the highest-speed processors, while not exposing the lower-speed processors to the OS. Such a configuration allows for reduced costs and power consumption, providing large-memory and top system performance. Similarly, the customer can mix and match I/O expansion options to fit their unique application needs, making it possible to deliver the industry’s most versatile and flexible high-end x86 systems. Coupled with the price and performance attributes, solutions based on vSMP Foundation provide customers the best value for their money.

To learn more about the vSMP Foundation architecture and solutions see the Whitepaper download page.

Back to top