Filling the Multicore I/O Gap on x86 ServersFilling the Multicore I/O Gap on x86 Servers
With the rise of hyperconverged systems, the opportunity to sell parallel I/O storage has never been better.
September 8, 2015
It has been apparent for some time now that the rapid pace at which x86 processors are being developed is creating a challenge for storage arrays to keep up. The typical x86 processor is now a multicore entity. The trouble is, most storage systems can’t contend with all the I/O requests that multicore processors can generate.
For that reason the folks at DataCore Software, a provider of storage virtualization software, is making a case for applying a parallel I/O architecture to x86 servers. While the concept of parallel I/O architectures is as old as the mainframe itself, use of parallel I/O architectures using x86 servers has been limited to high-performance computing (HPC) environments.
Augie Gonzalez, director of Product Marketing at DataCore, said when it comes to more traditional x86 servers, parallel I/O storage has become something of a lost art. With multiple virtual machines now running on multiple cores on x86 servers, Gonzalez said it’s time for solution providers to help revive that art to fill that gap that reliance of traditional serial I/O approaches to storage are now creating. For example, Gonzalez noted that the parallel I/O architecture developed by DataCore can deliver up to 400,000 IOPS in less than a millisecond that can be distributed across multiple cores. The end result, he said, is that two servers can now do the work of 10 servers in about one-fifth the time.
With the rise of hyperconverged systems, Gonzalez said the opportunity to sell parallel I/O storage has never been better. DataCore has developed reference architectures for Lenovo, Dell, Huawei, Fujitsu and Cisco (CSCO) servers that span all price points. In fact, with the rise of software-defined storage it’s becoming easier to implement complex storage architectures because they now are managed at a much higher level of abstraction.
DataCore has been making a case of a software-centric approach to storage in x86 servers for years. But it’s only just now that the issue is coming to fore as more IT organizations find themselves acquiring x86 servers that have multiple cores that wind up sitting idle most of the time. If the storage systems can support more cores on those servers, then it actually becomes a lot more feasible to consolidate the entire data center environment. Just as importantly, the solution provider that shows an IT organization how maximize all the cores running on their servers will most likely become one of their most trusted advisers for years to come.
About the Author(s)
You May Also Like