WekaIO announced support for Mellanox Technologies InfiniBand and Ethernet intelligent interconnect solutions for servers, storage and hyperconverged infrastructure. High performance computing, machine learning and technical compute workloads can achieve higher performance, scalability and lower latency by utilizing WekaIO software over Mellanox interconnect solutions.
Mellanox is the leading supplier of performance interconnect for high-performance GPU clusters used for deep learning workloads. WekaIO Matrix™ software coupled with ConnectX-series InfiniBand and Ethernet adapters enable performance improvements to data-hungry workloads without customers needing to modify existing networks.
Furthermore, WekaIO has developed an optimized zero-copy network software stack based on PCIe virtualization and Data Plane Development Kit (DPDK) technologies providing a 10x boost in packet processing for low latency, high throughput applications utilizing Mellanox ConnectX Ethernet and InfiniBand networking technologies.
"The combination of WekaIO software and Mellanox Ethernet and InfiniBand adapters provides our mutual customers with a high-performance file system solution that can address a broad range of applications," said Dror Goldenberg, vice president of architecture at Mellanox. "The WekaIO team has demonstrated performance improvements of 100 times compared to other file system options, leveraging the Mellanox interconnect technology and acceleration engines."
"New compute-intensive applications are generating diverse workloads that require a fundamental shift in the way storage systems are built to deliver on-demand performance at massive scale," said Liran Zvibel, co-founder and Chief Technology Officer at WekaIO. "The WekaIO Matrix solution now supports a broad portfolio of Mellanox Ethernet and InfiniBand platforms, providing businesses with HPC class performance and cloudlike scalability, without the need for network gateways."
A combined solution from WekaIO and Mellanox delivers all-flash performance and cloud scalability and economics.