GPU POWERed Stream Processing:
A New Level of Speed & Efficiency
While CPU-bound data processing solutions are limited to processing thousands or maybe millions of events per second, FASTDATA.io exploits the power of NVIDIA GPUs to give your business the ability to process billions of events per second on a single server.
Today, we are proud to announce the fastest benchmark to date for our GPU-native stream & hyper-batch processing software, Plasma Engine™, running on IBM POWER9.
Leveraging IBM® Power System™ Accelerated Compute Server, Plasma Engine v1.1.4 processed 1.6B rows/sec (or 27GB/s) in total throughput using a SparkSQL haversine query to calculate and compare distance between two GPS points on Earth.
Designed to fully exploit the capabilities of its NVIDIA GPUs -- 4x Telsa V100s in this case -- IBM’s POWER9 Server eliminates I/O bottlenecks and accelerates memory transfer across both GPUs and CPUs, enabling Plasma Engine to process data faster than ever before.
Built specifically for data-intensive workloads, the POWER9 CPU has state-of-the-art I/O subsystem technology, including next-generation NVIDIA NVLink 2, PCIe Gen4 and OpenCAPI, allowing Plasma Engine to scale linearly across multi-GPUs using its one-of-a-kind “shuffle” capability. The resulting performance was over two orders of magnitude greater than any CPU-bound stream processing software on the market today.
Plasma Engine utilizes IBM POWER and NVIDIA GPUs not only to increase speed and performance exponentially, but also to reduce total data processing costs by more than 50% while cutting power and space requirements in the data center by over 90%.
What could Plasma Engine on IBM POWER9 do for your business?
Contact us today at firstname.lastname@example.org to find out.
IBM® Power System™ Accelerated Compute Server (AC922)
CPU: 40cores @ 3.8Ghz
HDD: 2x1.92TB SSD
OS: Ubuntu 18.04