Qing Zheng Presents at Upcoming IEEE International Conference on Cluster Computing
New Mexico Consortium and Los Alamos National Laboratory (LANL) scientist, Qing Zheng, will be presenting his recent paper titled, KVCSD: A Hardware-Accelerated Key-Value Store for Data-Intensive Applications at the upcoming 2023 IEEE International Conference on Cluster Computing which will be held October 31-November 3, 2023, Santa Fe, New Mexico, USA.
Authors on this paper include Inhyuk Park, Soonyeal Yang, Woosuk Chung of SK hynix, Qing Zheng, Dominic Manno, Jason Lee, David Bonnie, Gary Grider of LANL, Bradley Settlemyer of NVIDIA, and Youngjae Kim of Sogang University.
What is the KV-CSD project? KV-CSD is a project where they are investigating new forms of storage namespaces that take advantage of emerging hardware-accelerated key-value stores to enable potentially orders of magnitude speedups for scientific data analytics on LANL’s future computing platforms.
More background: Popular software key-value stores such as LevelDB and RocksDB are often optimized for writing but they tend to also be good at reads. This is because that while data is initially stored in a write-optimized format, in the background it is asynchronously transformed into a format that better accommodates reads. Write-optimized key-value stores can still block writes. This happens when those background workers cannot keep up with the foreground insertion workload.
Their paper makes a case for a hardware-accelerated key-value store that allows running performance-critical operations — such as background data reorganization and queries — on storage rather than on a host. This better hides background work latency, prevents it from blocking foreground writes, and improves overall I/O efficiency.
Their prototype, called KVCSD, is a key-value based computational storage device consisting of an NVMe SSD and a System-on-a-Chip (SoC) that implements an ordered key-value store atop the SSD. Through offloaded processing, KVCSD streamlines data insertion, reduces host-device data movement for both background data reorganization and query processing, and shows up to 10.6x lower write times and up to 7.4x faster queries compared to the current state-of-the-art on a real-world scientific dataset.