ScaleFlux Delivers Ultra-Dense Computational Storage Solution for High-Performance Databases

RoF increases the capacity of a Redis cluster by utilizing high-performance Flash memory as a storage tier below main memory (RAM), and its performance is strongly influenced by data access patterns. The higher the Flash-to-RAM ratio, the more dominant the Flash storage layer becomes to database performance, and in many cases the highly intensive workload can wear out ordinary solid-state drives (SSDs). To address this challenge, users are required to either buy and deploy 2-3x the capacity of SSDs that their database really needs, or replace worn out SSDs after as little as 18 months.

The combination of the CSD 2000 with Supermicro’s FatTwin creates a compelling solution to this problem. CSD 2000 meets the endurance and performance needs of RoF workloads and FatTwin delivers high-density compute through its half-width design. Benchmarking data shows consistent, low-latency performance with RAM hit ratios as low as 60% using just two CSD 2000 devices per node. The half-width architecture of the FatTwin platform slashes physical space requirements, while transparent datapath compression featured in the CSD 2000 cuts the storage space requirements — all while addressing the key workload concerns of a RoF deployment: write endurance and read performance.

"The small number of SSDs required by RoF deployments leads to some critical design considerations, as users require SSDs to individually offer high endurance mixed with excellent workload performance," said Tong Zhang, co-founder and chief scientist of ScaleFlux. "Our collaboration with Supermicro addresses these needs and appeals to a variety of use cases. We look forward to astounding customers with the performance and cost efficiency."

Key benefits of CSD 2000 with FatTwin for RoF deployments include:

• Data center optimization: RoF only requires a small number of SSDs to achieve the needed capacity, FatTwin is designed to enable two nodes per rack unit, doubling the compute density compared to full-width systems.
• Aligning service intervals and reducing total cost of ownership (TCO): The compression capability in the CSD 2000 enables SSDs to last three years or more in the write intensive RoF environment.

"We’re excited to collaborate with ScaleFlux and other ecosystem players to enable solutions for improving data center density and application performance," said Vik Malyala, senior vice president, Supermicro. "As a global leader in high performance, high-efficiency server technology, and innovation, our goal is to work with industry leaders offering leading computational storage solutions."

The ScaleFlux® Computational Storage Drive CSD 2000 Series brings exceptional performance, scalability, and TCO savings to mainstream flash deployments. ScaleFlux drives combine up to 8TB of the latest 3D NAND Flash technology with hardware-accelerated compute engines, achieving incredible data read/write speeds and consistent low latency.

"This high performance, innovative combination of ScaleFlux computational storage and high-density next generation systems based on Supermicro FatTwin server platforms clearly delivers enhanced scalability and performance for Redis," said Joel Wineland, CTO at Racklive, data center division of ASA Computers, who provided resources for this testing. "We expect to see many customers in a variety of verticals including Ad tech benefit from such implementation. Looking forward to incorporating these elements into our portfolio of rack-level roll in solutions to maximize value and enhance performance." Preconfigured RoF nodes are now available on Racklive. Read the benchmarking white paper to learn more.

‍