The focus of ngGrid is to bring the power of a new class of processor to a wider market. Previously CPU-bound applications can be given a massive performance boost by converting the core routine (for example a Monte Carlo simulation or a credit default engine) to run on a co-processor residing on an expansion card in the computer. The incredible power is achieved by converting the algorithm effectively into a hardware design, which then can leverage significant innate parallelisms in the processor.
The co-processors we currently use are Field Programmable Gate Arrays - FPGAs. These are a commodity device used in telecoms equipment, military systems and by the scientific community (eg NASA, genetics research, geological modeling). While the devices themselves are mature (and improving in speed and capacity faster than RISC/CISC CPUs -- ie 'normal' processors -- we are the leading provider of FPGA solutions in the City.
We can help you perform the up-front analysis to identify candidate algorithms. While most calculations can be translated, only a subset will be cost effective. Ideal candidates are low to medium context (eg limited historical data) that are principally CPU bound (ie not limited by database or network bottlenecks).
We can translate the chosen calculations into super-efficient designs that run on the FPGA. While compilers exist that will translate algorithms into FPGA designs from modified C code, the true potential of these devices is not exploited unless the underlying architecture is incorporated into the design. We can provide the required expertise.
We can integrate the solution into your existing architecture and infrastructure. Our City experience has taught us that having an incredible standalone system is not enough -- it must be usable and integrated into existing infrastructures and processes. Whether it's integration with an existing grid farm or providing state of the art web services or SOA integration, we have the expertise to deliver end-to-end successful implementations.
The incredible performance benefits of this technology mean that those companies that seize the opportunity to exploit it, will gain an advantage over the competition. The extra capability gained can be used in a number of key ways:
You can increase the scope and number of scenarios to gain a more comprehensive view of your risk.
For those heavy-weight calculations that you can only update overnight due to the compute time required, you can now have them as real-time.
Rather than working with stock parameters, you can easily explore more options with complex products without waiting for minutes or hours for each scenario.
This lets you reduce pressure on support cycles and reduces operational risk due to failed batches -- you can re-run if needed.
Contact us today to find out how we can help.