Complex multi-tier applications and components communicate with one another using the network infrastructure in the datacenter and between datacenters. Often, these communications paths are the bottleneck that reduces both scalability and application performance. Many network virtualization suppliers have emerged to address this challenge. I had a chance to speak with Raj Kanaya, CEO and Co-Founder of Infineta, about an interesting innovation that would certainly offer the ability to pump more data through these communications links.
Over the years data compression and data de-duplication have been applied to the stream of data flowing over the network. This reduction in the amount of data required would optimize performance and reliability of that data link. Data compression is encoding data show that fewer bits are needed to transmit that data from here to there. Data de-duplication (called data dedupe in the industry) is looking for replicated blocks of data and sending only a single copy to reduce the network traffic.
Until very recently, industry standard processors (X86) have been used in the network appliances and network servers deployed by most organizations. While this approach has resulted in a significant increase in network performance, these tools were limited by the fact that one section of the data stream at a time could be processed.
The folks at Infineta looked at the problem and thought of a different approach. They developed custom microprocessors that were designed to apply parallel processing techniques to this data stream so that more data could be pumped through the system in a shorter period of time.
While it is very likely that this approach will be too expensive for smaller organizations, large organizations pumping huge amounts of data around their datacenters or into and out of their cloud applications will find Infintea's technology very helpful. The benchmark data I was offered was very impressive. Obviously benchmarks are artificial and may not properly represent what your organization is doing. It would be good to try this technology out in your own environment.
If your organization has a need for a significant improvement in data throughput, it would be worth the time to chat with Infineta about their products.