Intel Atom Processor Offers Storage Value
Using a low-cost Atom processor from Intel in a storage system can lower costs, improve power efficiency, and take up less rack space while also providing acceptable performance.
April 29, 2011
The processing power of a storage system becomes increasingly important as we ask storage systems to scale larger, process requests faster, and enable more capabilities. The focus for the most part has been on putting the most powerful processor possible in a storage system so it can meet those challenges. This approach remains the best course of action for conventional dual-controller scale-up systems, but for scale-out storage systems there may be another alternative --Intel's Atom.
Why would you want to put a processor designed essentially for the netbook market into a storage system? The easy answer is when you don't need the performance of a high-end CPU like the Westmere processor. Processing performance that goes unused adds to the cost of a storage system, and the user can't derive any benefit from that expense. An example of where processing power may go to waste is in scale-out storage systems that are either focused on the small to midsize enterprise as we discuss in our article on mid-range scale-out NAS or high-node-count scale-out systems as we discussed in "Building Affordable Scalable Storage Infrastructures". In both of these cases, the nature of scale-out storage when using standard high-powered processors can leave processing power underutilized and overpaid for.
Scale-out storage in essence aggregates multiple storage servers (nodes) into a single cluster that looks to the attaching users and servers like a single storage entity. The idea is that as storage capacity demands increase, more storage nodes are added to the system. Each one of these nodes has storage capacity, additional bandwidth, and additional processing power, which are all aggregated together with the other nodes so that performance scales as capacity is added. The challenge is that much of the processing performance sometimes goes unused when scale-out storage is used for basic file sharing or when scale-out storage is used in high-node-count configurations like you would find in online applications, online storage, and long-term archive environments. Too much processing power is almost as bad as too little processing power in that you are paying for something that you are likely never going to be able to take advantage of.
In fairness to the scale-out storage manufacturers, there were simply not a lot of alternatives in processor selection. Now though, with the advent of Intel's Atom processor, there is an alternative. While the processor does not perform as well as the processors typically found in servers, as I explained above, in scale-out storage that may not be needed. If acceptable performance can be achieved for the intended market, the Atom can be a perfect match for certain scale-out environments. Not only does using an Atom processor keep the percentage of CPU utilization at a higher level, these processors also require less power and generate less heat, which should drive down the cost of operation even further. That also means that these systems could be placed in denser packaging, helping the other area of concern within the data center--floor space.
Switching to the Atom processor means that scale-out storage systems can lower costs, improve power efficiency, and take up less rack space--all things that are sometimes a roadblock for today's scale-out storage systems. It is fair to assume that scale-out storage systems using top-of-the-line processors will still be needed, especially in primary storage applications where IOPS is as important as capacity. But for long term data retention or for basic file sharing, using the Atom in scale out storage has a lot of value.
Follow Storage Switzerland on Twitter
George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Storage Switzerland's disclosure statement.
About the Author
You May Also Like