Flash Memory Summit: NeuroBlade Demonstrates Acceleration of Analytics Workload by Up to 100x Using SQL Processing Unit Technology

NeuroBlade, Inc., in data analytics acceleration, is changing the speed and performance of query processing on petabyte-sized datasets.

Hardware Enhanced Query System (HEQS)

It announced the demo of its NeuroBlade Hardware Enhanced Query System (HEQS) utilizing Kioxia Corp.’s CM7 Series Enterprise NVMe SSDs. This solution empowers customers to harness the full potential of high-performance SSD throughput, resulting in query performance improvements of up to 100x.

Despite improvements, general-purpose CPUs struggle to keep up with the growing I/O throughput, which has increased by approximately 17x in the past 5 years. The company’s purpose-built SPU architecture overcomes traditional CPU limitations, eliminating system bottlenecks and enabling high-speed query processing regardless of complexity. This delivers performance, reduces footprint, and improves energy efficiency. With the firm’s solution, organizations can leverage their data assets to identify new business opportunities and deliver enhanced services at a lower cost.

SQL Processing Unit (SPU)

The HEQS is a rackmounted server that combines the company’s SQL Processing Unit (SPU) with high-bandwidth PCIe Gen 4.0/5.0 NVMe flash storage from companies like Kioxia, and their CM7 Series SSDs. By processing data closer to -storage, the HEQS maximizes the performance of NVMe flash drives, minimizing data movement and accelerating overall system performance for data-intensive tasks. This enables customers to achieve pure linear scalability, scaling their performance to match their largest requirements.

“Our collaboration with Kioxia strengthens our go-to-market strategy and underscores the growing demand for innovative and powerful solutions,” said Elad Sity, CEO and co-founder. “This breakthrough empowers organizations to keep pace with their exponential data growth, elevate analytics performance, and gain an invaluable competitive advantage. We owe this success to our exceptional engineering team, who have worked closely with companies like Kioxia to unlock this level of innovation for data analytics.”

The HEQS integrates into a customer’s environment, accelerating analytics workloads by 10-100x while reducing TCO by up to 10x and power usage by up to 85%. It connects to various database query engines without requiring changes to existing data, queries, or code. Thisenhances performance for analytics workloads such as business intelligence (BI), data warehouses, data lakes, ETL, and more, enabling organizations to achieve substantial improvements in performance, cost, and power efficiency.

“We welcome the collaboration with NeuroBlade and their efforts to address the performance gap in data analysis” said Neville Ichhaporia, SVP and GM, SSD business unit, Kioxia America, Inc. “Our CM7 Series SSDs are well-suited for data analytics workloads that require high throughput access to storage. We are delighted to work with NeuroBlade and contribute to the delivery of an innovative solution that accelerates widely used database applications.”

The demo includes HEQS G100, featuring 32xCM7 Series NVMe SSD drives, 512GB memory, and 8xNeuroBlade SPUs. By benchmarking standard TCPH tests with a 10TB scale factor, the demo highlighted the results achieved by the NeuroBlade accelerated solution compared to a standard server running software only, equipped with 128 CPU cores. The benefits of the firm’s solution included up to 80% utilization of Kioxia SSD drives I/O throughput, up to 23x query performance improvement, and up to 21x improvement in consistent Query Data Scan Rate (QDSR).

The firm’s demo will be displayed at the Kioxia booth during the Flash Memory Summit 2023, taking place from August 8 to 10, in Santa Clara, CA.