Follow Datanami:
November 9, 2011

Solarflare Partnership Takes Aim at I/O

Datanami Staff

High-performance networking companies are taking aim at the growing big data trend with a wealth of new solutions set to tackle the I/O problems that plague many data-intensive applications. For instance, recall that just yesterday, Mellanox announced their answer to big data performance woes with UDA 2.0.

Today Solarflare, a company that specializes in 10GbE network hardware and software, announced a partnership with the data decentralization company, Fusion-io, to decrease latency and boost performance of data-intensive applications.

Solarflare says that their 10GbE products, when matched with Fusion-io’s create a win-win for big data applications. According to the release today, “advancements in data search and analytics, web object caching and search, and data retrieval provided by Hadoop/HDFS, MapR, Memcached and PCIe-based flash, customer applications are now inhibited by network I/O performance.”

The networking company calls their 10GbE offerings “application-intelligent” and is staking a claim in the low-latency for big data camp, stating that they have the highest-throughput Linux kernel and Windows performance. They point their server adapters’ ability to handle the latest 2/3 smart flow management tools that let them “intelligently” match network performance via the server based on customer requirements. The addition of the Fusion-io partnership seems aimed at solving the latency puzzle by taking care of both the network and storage memory sides of the equation.

Fusion-io’s ioMemory solution can whisk immediately-needed data into the servers where the application processing occurs to take care of the latency concerns while Solarflare’s 10GbE products take care of similar issues on the network I/O side. As referred to above, Fusion-io calls itself a “data decentralization” company, which refers to their approach to storage memory that moves process-critical data closer to the CPU for processing to reduce latency.

Fusion-io’s io Memory modules are designed fast processing of big data workloads. They claim that “one ioMemory module contains 100 times the capacity density of dynamic random access memory, or DRAM, to provide applications fully scalable memory for accelerating throughput and reducing latency.”

According to Solarflare’s Bruce Tolley, “with the growth of big data applications and customer adoption of Fusion-io PCIe flash products, the network has become the bottleneck. By accelerating network I/O, increasing throughput and decreasing latency, Solarflare unlocks the processing power of today’s servers for analytics.”

Datanami