tyronefoundations
Home Products High Performance Computing SMP Solutions
SMP Solutions See Related Products:
Contact Us @
Bookmark and Share
HPC applications have always needed faster and larger systems, historically served by SMP systems. Symmetric Multi Processors (SMPs) enable multiple CPUs in a network share the same board, memory, I/O, and operating system. The dominance of SMP was due to their ease of deployment, management, and programming capabilities. However, these fully proprietary systems were expensive in that they:
+ Use custom processors, custom chipsets, and ASICs to create high-speed backplanes
+ Work on custom Operating Systems (OS)
+ Require significant investment in system-level architecture by computer manufacturers.
Also, most of the SMP systems with 8 processors or more, utilize non-x86 processors, thus making them very expensive.

Commodity server clusters connected with high-speed interconnects offer a cheaper alternative to SMP systems. However, although these cluster solutions deliver more performance than traditional SMPs, they are more difficult to program, implement, and manage:
+ Clusters require network boot or other such centralized OS deployment techniques
+ Specific programming model is needed to accommodate the distributed computing resource
+ Each processor can access only the cluster node's limited local memory.

Tyrone’s SMP solution combines the ease-of-use of SMP systems and the price advantage of commodity clusters by leveraging off-the-shelf x86 components. This innovative architecture aggregates multiple x86 systems into a single virtual x86 system to deliver an industry-standard, high-end SMP computer with up to 128 processor cores and memory of up to 1 terabyte - without modifying the OS or the application.

Tyrone’s SMP system:
+ Offers a revolutionary computing paradigm by using software to replace custom hardware and components
+ Aggregates the compute, memory, and I/O capabilities to presents a unified virtual system for the OS and software    applications
+ Enables aggregation of memory bandwidth across the board (as opposed to the traditional SMP architecture where    memory bandwidth decreases as the machine scales
+ Offers a single system image with fewer nodes, hence reducing the complexities involved in managing multiple OS,    cluster file systems, application provisioning, and system interconnects
+ Can run any type of HPC application be it distributed applications or large memory footprint jobs on a shared    infrastructure
+ Offers the power of SMP systems at the price of commodity clusters. Faster simulations also lead to significant cost    savings.

Problem Statement Realtive Cost
SMP x86 clusters Tyrone SMP
R&D and manufacturing cost High Low Low
Initial acquisition cost
Installation and ongoing cost Low High Low
Programming cost

Scaling(CPU Cores) Up to 960 parallel processing cores
Peak Performance Nearly 64 in single precision and 5.536 teraflops in double precision using 35U rack space
Network/ Interconnect Compatibility InfiniBand and Ethernet


The following are some applications on which the tyrone's SMP has been tested:
CAE Life Sciences Numerical Simulations
  • ANSYS Mechanical
  • ABAQUS/Explicit & Standard
  • LSTC-LS-DYNA
  • FLUENT
  • ANSYS CFX
  • CD-adapcpStar-CD
  • AVF FIRE
  • Gaussian
  • Amber
  • GROMACS
  • Schrodinger jaguar
  • Schrodinger Glide
  • NAMB-DOCK
  • GAMESS
  • GOLD
  • mpiBLAST
  • BLAST
  • MOLPRO
  • OpenEyeFred
  • OpenEyeOmega
  • SCM ADF
  • HMMER
  • The Mathworks, MATLAB
  • Interactive Supercomputing, Star P
  • Revolution Computing R



 


© 2009-2019 Tyrone Systems All Rights Reserved.