High Performance Computing Solutions

Netweb's HPCC solutions are built using leading industry-standard building blocks and best-in-class partner products. The result is a highly scalable HPCC solution that can deliver exceptional performance and manageability at a fraction of the cost of competing solutions. We offer a wide choice of systems and interconnects allowing you to choose the best solution for your requirements. Our systems range from single socket to quad-socket and in various form-factors including the revolutionary Twin1U which packs two servers inside a single 1U chassis to deliver unparalleled density. This 0.5U density (84 server nodes in a standard 42U rack) combined with optional onboard InfiniBand or 10GbE makes these platforms an excellent choice for high-performance computing (HPC) clusters where space, cost, energy-efficiency and density are high priorities.

Our 10-blade server system based on the SBI-7125C-T3 blade is an extremely cost-effective supercomputer. This solution now provides the scientific community the opportunity to enable a personal supercomputer next to their desks with the same computing power previously only available via large server installations in a dedicated IT room. This optimized blade solution features 93% power supply efficiency, innovative and highly efficient thermal and cooling system designs, and industry-leading system performance-per-watt (300+ GFLOPS/kW), making it the greenest, most power-saving blade solution.

Large Memory SMP systems

HPC applications have always needed faster and larger systems, historically served by SMP systems. The dominance of SMP was due to the ease of deployment, management and programming. However, these fully proprietary systems were expensive as they used custom processors, custom chipsets and ASICs to create high-speed backplanes as well as custom Operating Systems (OS).SMP systems require significant investment in system-level architecture by computer manufacturers as well as substantial amount of development time.The result of the high R&D investment is an expensive solution that uses proprietary technology based on custom hardware and components. To date, most of the SMP systems with 8 processors or more utilize non-x86 processors, which is a large contributor to the high price of SMP systems. In addition to the high price, these systems generally "lock" the end customer into a specific computer manufacturer due to its proprietary hardware architecture and software stack.

An alternative for large SMP systems was the adoption of commodity server clusters connected with high-speed interconnects. Cluster solutions deliver more performance than traditional SMPs, but are more difficult to program, implement and manage. It leads to higher OS deployment cost and complexity such as network boot or other centralized OS deployment techniques are required, resulting in a need for higher IT skill sets. Providing the application with access to common storage requires use of cluster file-system, SAN or NAS deployments. In addition to complexity, cluster deployment poses two challenges at the application level. i) Specific programming model is needed to accommodate the distributed nature of the computing resource ii) Each processor can access only the "cluster" node's local memory, which is usually limited. Netweb delivers a software-based computing-architecture that combines the advantages of shared-memory systems and the price points of commodity clusters by leveraging off-the-shelf x86 components.

This innovative architecture aggregates multiple x86 systems into a single virtual x86 system, delivering an industry-standard, high-end SMP computer with up to 128 processor cores and up to 1TB memory - without modifying the OS or the application. This solution uses software to replace custom hardware and components, to offer a new, revolutionary computing paradigm. It aggregates the compute, memory and I/O capabilities of each system and presents a unified virtual system to both the Operating System and the applications running above the OS.

It enables the aggregation of memory-bandwidth across boards, as opposed to traditional SMP architecture where memory bandwidth decreases as the machine scales. This enables our solutions to show close to linear memory bandwidth scaling. It can run any type of HPC applications providing best of breed performance for both cluster and SMP applications.