Integration of NeuroBlade’s SQL Processing Unit (SPUTM) technology into Velox introduces unprecedented performance advancements (30x), ushering in a new era of data analytics.
Palo Alto, Calif. — NeuroBlade, a pioneering leader in data analytics acceleration, works with the Velox community to open source hardware (HW) acceleration to the Velox execution engine. The seamless integration of NeuroBlade’s cutting-edge SPU technology into Velox yields a remarkable acceleration of over 10 times, on top of the 3 times performance improvement already achieved by Velox using software optimizations. This effort aims to empower organizations to process massive datasets efficiently, setting new benchmarks in performance and efficiency.
“We have created Velox to unify software-based optimizations into a single execution engine that can be reused across engines such as Presto, Apache Spark, and beyond. Our long-term strategy for Velox, however, is to provide a unified framework that allows hardware accelerators to be seamlessly integrated into data management”, commented Pedro Pedreira, Software Engineer and Velox lead at Meta. “We are optimistic to see NeuroBlade’s hardware acceleration performance improvements coupled with our open source software initiatives. This result serves as a validation of this model, and lays the groundwork for further acceleration initiatives”.
Elad Sity, CEO and Co-founder of NeuroBlade, highlighted the significance of the shared effort, “Collaborating with the Velox community to open source hardware acceleration for Velox highlights our dedication to innovation in providing high-performance and efficient solutions. Early results are promising, showcasing an acceleration of over 10 times, on top of the 3x achieved by Velox thus far through software optimizations. This effort heralds an era where organizations can adeptly manage escalating data volumes, elevate analytics performance, and gain a significant competitive edge.”
Velox, a state-of-the-art open source unified execution engine, consolidates various software-based optimizations from the query engine ecosystem into a singular high-performance library. Designed for seamless integration with various query engines, starting with Presto and Apache Spark, Velox provides a robust foundation. However, conventional CPUs struggle to efficiently manage analytical queries surpassing 2-3GBps due to limitations in data retrieval and query complexity. NeuroBlade’s SPU technology overcomes these challenges by creating a dedicated analytics processor that achieves an optimal storage-to-compute balance. NeuroBlade’s SPU architecture offloads the CPU by efficiently executing many sub-query operations and doing the data reduction close to storage. Further, the SPU is designed to execute queries at a constant throughput, hence eliminating query complexity as a source of variable latency.
The work with Velox includes the seamless integration of NeuroBlade’s SPU hardware acceleration into Velox. This is done through novel Velox APIs that allow arbitrary parts of query execution to be offloaded to an accelerator.
The shared effort of NeuroBlade and Velox introduces exciting prospects in data analytics acceleration solutions. The work marks the pioneering integration of NeuroBlade’s cutting-edge SQL Processing Unit (SPU) technology with the Velox execution engine, a first in the market. By synergizing NeuroBlade’s hardware acceleration capabilities with Velox’s steadfast dedication to open source, this effort effectively caters to the escalating need for performant and efficient data analytics solutions. Anticipated for release later this year, the implementation will be accompanied by a detailed white paper describing the methodology, benchmark techniques, and outcomes achieved through this innovative approach.
NeuroBlade will showcase its technology at the Big Data Analytics LDN (London) conference on September 20-21 in booth #216. Click here for more information about NeuroBlade’s participation at the show.
NeuroBlade is pioneering the new standard for data analytics acceleration, enabling organizations to achieve exceptional performance and efficiency in processing large-scale datasets. NeuroBlade’s SQL Processing Unit (SPU™) technology and Hardware Enhanced Query System (HEQS™) revolutionize data analytics by delivering accelerated query performance and unlocking valuable insights in record time. For more information, please visit www.neuroblade.ai .