Big Data / GPU Processing
FlexSystem moved into in-memory processing in 2011 and has continued to pioneer in providing fast access to data. Benefits of faster access means all types of reporting are quicker especially across applications which slow down as a result of the involved complexity.
FlexSystem sees Big Data as the data within the enterprise. Often users are frustrated by slow access to data even with today’s volumes but as data volumes increase and as geographical barriers continue to be knocked down more processing power will be required for reporting purposes. The ability to do more relevant reporting across corporate data will require even more processing power. Even a 50 times speed increment means that a report taking 50 seconds will be done almost instantaneously and becomes very usable.
What is GPU (Graphics Processing Unit) processing? A typical computer has the CPU within one single physical component and typically contains 4 to 8 cores, each core being a CPU under the command of the main processor that can run on a concurrent basis. The GPU contains thousands of cores (example Nvidia Tela K40 has 2,880 cores and utilises Kepler Architecture) and is massively parallel allowing the CPU and GPU to crunch through massive volumes of computations at high speed.
FlexSystem General Purpose GPU processing (GPGPU) has been designed to utilise both the CPU and GPU for processing where the main processing CPU offloads to the GPU both CPU intensive and time intensive processes thus freeing up the CPU to crunch numbers and assign tasks to other parts of the computer and in so doing allows the applications to run significantly faster. Significant changes were made and continue to be made by Nividia to allow these GPU’s to be accessed and addressed by programmers and the resultant language to enable this process was called CUDA. Also see Visualization.