A Data Center without Hard Drives: Stanford's RAM Plan
Stanford researchers would like to replace hard drive storage and put everything in computer memory.
Memory is in, hard drives are out. If Computer researchers at Stanford University have their way, they'd eliminate hard drive storage for their computer system and store everything in computer memory. They are calling it the RAMCloud, proving once again that you can spice up anything with the word “cloud” whether it has anything to do with cloud networks or not.
The researchers argue that although hard drive storage capacity has increased dramatically over the last four decades, the device's performance hasn’t increased as much and has fallen behind the needs of large-scale Web applications. Many solutions have been proposed, including the use of solid-state drives, but the folks at Stanford have proposed a whole new class of storage that uses dynamic random-access memory exclusively.
One of DRAM's benefits is its lower latency -- lower, in fact, than any other type of storage. It would have other benefits as well, such as quick recovery. In another paper, the Stanford researchers claim that an average-sized server using RAMCloud could recover from a crash in 1.6 seconds.
Of course, there are two major potential problems with using DRAM. One is the cost. DRAM has a much higher cost per data unit than most other types of storage available. However, because this storage class would primarily be used in huge Web-based applications (such as retail), they could probably afford to splurge a bit.
The second issue is the clincher. DRAM is designed to work when it has power supplied to it. Whatever is stored on it will quickly degrade without power, so to use it in this capacity will require a huge amount of uninterruptible power. That means that if the worst actually comes to pass and the data center is left powerless for an extended time period, the center's employees will need a copy of their data on more conventional media so that they can restore it to the DRAM systems. Network centers have become adept at maintaining constant power level for some time now, so this wouldn’t be anything new.
All in all, it’s an interesting idea that could dramatically improve the performance of scalable systems.
Greg Crowe is the associate technology analyst for the Government Computer News lab.