RAMcloud Research Project Relies on DRAM for Primary Storage

One of the more potentially revolutionary technologies that will have a profound impact on the enterprise in 2013 is the rise of in-memory computing. While most everyone agrees that more database and application logic will be running in-memory in 2013, no one is quite sure what the impact on disk-based storage systems is going to be.

There are those that argue that disk-based storage systems will always be needed to insure the availability and integrity of data, given the volatile nature of memory. But there is also a cadre of researchers working diligently to eliminate the need for primary disk storage altogether.

One of the more promising projects pursuing that line of research is known as RAMcloud. Led by researchers at Stanford University, the goal with RAMcloud is nothing short of creating a general-purpose storage system that would run completely in DRAM on server processors. As the cost of DRAM continues to drop, while the amount of sheer processing power increases, thanks to the advent of multicore processors, it’s become a lot more feasible to think of DRAM as a primary storage medium.

In a recent webcast hosted by the Cloud Advisory Council, Stanford professor John Ousterhout details not only how RAMcloud works, but how it will be used to overcome a wide range of latency issues that currently inhibit the development of applications involving the manipulation of petabytes of data in real time.

It’s hard to say at this juncture whether RAMcloud will ever show up as production system in a data center. But it clearly shows the direction in which data center infrastructure technology is heading, starting with a coming sea change in the way storage is managed. Of course, it’s not likely that all this will come to fruition in 2013. But the one thing that is for certain is that from an enterprise IT perspective, RAMcloud may be one of the most interesting nascent data center technologies to come down the pike in well, pun intended, recent memory.