Rambus Partners with Los Alamos National Laboratory on Smart Data Acceleration.Deployed at LANL, the SDA platform is designed to optimize the performance of in-memory databases, graph analytics and other Big Data applications.
“With the advent of new tiers of the memory hierarchy coupled with high-speed modern networks that are being made available to the HPC community, independent scaling of memory resources in different tiers may be possible, which will help economically match the needs of future applications and workflows of extreme high-end HPC,” said Gary Grider, leader of the High Performance Computing Division, Los Alamos National Laboratory.
“Our initial tests of the Rambus SDA platform show that its performance is well-matched to HPC interconnects, and given the ability to process information close to memory, it can enable the reduction of data movement through efficient workflows. I think the win for large scale simulation is in this combination of independently scalable, high-performance, high-durability, large capacity and processing near memory.”
Addressing major issues facing servers and data centers in the age of Big Data, the SDA Research Program targets significant improvements in the acceleration and offload of computation as the industry explores new paradigms such as near data processing. In addition to compute offload and acceleration with flexible processing engines to enhance system performance, key focus areas include minimizing data movement and leveraging the improved bandwidth, and reducing latency of DRAM and other memory technologies.
“LANL is an ideal partner for our SDA Research Program as we collaborate with experts to validate our approach and overall direction,” said Laura Stark, senior vice president and general manager of the Emerging Solutions division at Rambus. “Our goal is to provide significant improvements in the performance of data-intensive applications through optimization of both software and hardware for compute acceleration and offload.”
The Rambus SDA Research Program provides a platform to investigate near data processing system architectures that include software, firmware, FPGAs and large amounts of memory. The platform can be used to test new methods to optimize and accelerate analytics for large data sets, including in-memory databases, financial services, ad serving, real-time risk analytics, imaging, transcoding and genome mapping among other applications.