Los Alamos National Laboratory and SK hynix will demonstrate recent enhancements to the ordered Key Value Store Computational Storage Device (KV-CSD) at the 2023 Flash Memory Summit. The project is a follow-on of a collaboration between Los Alamos’ High Performance Computing division and SK hynix that debuted at the 2022 Flash Memory Summit. The secondary index functionality added to the system’s capabilities has reduced data retrieval and offered speedups by allowing an index to extend beyond the primary key into the value.
“Often in large-scale scientific simulation, data is highly dimensional, even beyond the four dimensions of 3D space and time,” said Gary Grider, High Performance Computing division leader at Los Alamos. “Adding secondary indexing capabilities to the first-of-its-kind ordered KV-CSD device provides a way to approach data analytics of these highly dimensional simulation data. This capability will save time and energy in the understanding of the output of massive-scale computational science.”
The initial KV-CSD implementation enabled the Laboratory to analyze the benefits of a hardware-accelerated index at the storage device to serve large-scale, record-based simulation input-output. This indexing capability provided significant speedups due to a massive reduction in amount of data retrieved during a scientific data analysis process. To support rich query types, gain additional data reduction and build on the work to support ordered point and range queries, Laboratory researchers and SK hynix have enhanced the device to provide secondary index capability.
In what was a first for the nonvolatile memory express industry, last year the partnership succeeded in pairing indexing capability on the KV-CSD with Laboratory security science applications. The secondary index capability adds functionality for scientists to index specific data fields that make up a value, thus increasing performance and further improving support for highly selective queries.
“By combining Los Alamos’ renowned expertise in high performance computing and SK hynix’s technical excellence in storage solutions, we were able to deliver a state-of-the-art Key Value Store Computational Storage Device,” said Hoshik Kim, vice president and fellow of Memory Forest x&D at SK hynix. “This advancement will significantly accelerate the performance of data analytics, which is critical in various fields of data-intensive scientific computing such as artificial intelligence, genomics, climate modeling and more. SK hynix is committed to driving innovation in the way data is stored and processed by fostering collaborative efforts with industry ecosystem partners.”
Los Alamos National Laboratory and SK hynix have a memorandum of understanding toward the design, implementation and evaluation of the KV-CSD.
LA-UR-23-29094