Because storage is hardware, IT organizations are conditioned to think that the technology that manages storage should reside on a specific piece of hardware dedicated to that task. But as general-purpose processors become increasingly faster, it’s become a lot more feasible to run the storage system management software on the server.
That’s the thinking behind the approach that ScaleIO, which just raised another $12 million in financing this week, is taking toward the management of local disks as a virtual resource, which, as far as the application is concerned, appears to be just another storage area network.
According to ScaleIO CEO Boaz Palgi, Elastic Converged Storage not only eliminates the dependency on expensive SAN platforms, it makes it easier to take advantage of inexpensive disks and Moore’s Law to affordably scale I/O performance. The reason for this, says Pagli, is that rather than processing I/O serially, those I/O operations can now be processed in parallel using multiple cores on the server. This allows IT organizations to use disks and multicore processors to linearly scale application performance without having to invest in expensive Flash memory or in-memory computing platforms, says Palgi.
The end result, says Palgi, is not only better performance, but an actual 80 percent reduction in the total cost of storage using what he refers to as “virtual storage controllers.”
The real challenge traditional SAN platforms face is that upgrading controllers to keep pace with advances in server processors is time-consuming and expensive. As a result, there winds up being an imbalance between the processing capability of the server and the SAN systems attached to it. ScaleIO is arguing that rather than continuing to use a traditional SAN architecture, the time has come to rely more on the multicore processing power on the server that gets faster with each successive generation of servers.
That approach, adds Palgi, also makes it easier to manage storage resources at a higher level of abstraction, which Pagli refers to as software-defined storage.
At a fairly rapid clip, data center architectures are being transformed before our eyes. That means all the assumptions about how to best manage I/O don’t necessarily apply for every application, which means that all those SAN systems in the data center will possibly free up a lot of extra space, and also significantly reduce the amount of power consumed by orders of magnitude.
