Guest blogger: DCIG's Jerome Wendt on The Future of RAID

November 1, 2010 -- There was a sizable uptick last month in the readership of a blog entry that appeared nearly two years ago on DCIG's website on the topic of data loss on SATA storage systems. While this blog entry received a fair amount of interest when it was first published, exactly what prompted a resurgence of interest in this topic last month is unclear.

Maybe it's just an anomaly driven by the whimsical interests of Internet users who are for whatever reason searching on this topic.
However, it may be a more ominous indication that SATA disk drives are wearing out and that the traditional RAID technologies used to protect them are failing. As a result, users are looking for information as to why RAID, in some circumstances, is not doing the job in their environments.

The death of RAID (or at least RAID 5) has previously been forecast by some analysts. But even now, when I look at the features of new storage arrays, the number of RAID options that they support is always prominently mentioned.

A good example is Overland Storage's recent introduction of the SnapSAN S1000. It offers at least 10 different ways that RAID can be configured (including RAID 5) on a storage array that starts at less than $10,000, so it's clear that RAID is not dead or even on its last legs.
But there is no disputing that capacities of SATA disk drives will cross the 4TB, 8TB, 16TB and 32TB thresholds over the next decade. As that occurs, it becomes questionable whether current RAID technologies will be able to protect disk drives of these sizes. If the increased interest in DCIG's 2008 blog entry (see "Data Loss on SATA-based Storage Systems -- Coming Soon to Your Company?") is any indication, the answer would apprear to be 'no.'

Am I predicting the death of RAID? Clearly not. RAID technology is as much a part of the storage landscape as tape, and odds are that innovation will continue to occur in RAID to make it a relevant technology for the foreseeable future.

Yet it was clear from speaking to a few users and storage providers at Storage Networking World (SNW) last month that new approaches to protecting data stored on larger capacity SATA disk drives are going to be needed.
One specific company that I met with at SNW was Amplidata, which is already innovating in this space to overcome two of the better known limitations of RAID, including:

•The increasing length of time required to rebuild larger capacity drives. Rebuild times for 2TB drives are already known to take four hours or longer to complete, and in some cases -- depending on how busy the storage system is -- it can take days for a rebuild.
•The need to keep all disks in a RAID group spinning so no power savings can be realized. Spin down is likely to become more important in the years to come as more data is archived to disk, with it likely becoming a function of the storage array to intelligently manage and place the archived data on these drives -- as opposed to the software -- to facilitate the spin down of drives.

Amplidata's AmpliStor distributes and stores data redundantly across a large number of disks. The algorithm that AmpliStor uses first puts the data into an object and then stores the data across multiple disks in the AmpliStor system. By storing the data as an object, Amplidata can reconstruct the original data from any of the disks on which the data within the object resides.

This technique eliminates growing concerns about the rebuild times associated with large disk drives since the original data can be retrieved and reconstructed even if one, two or even more disks fail. Also, should disk drives in the system be spun down to save energy, they do not need to be spun up to retrieve needed data since the data can be retrieved and reconstructed from other spinning disks in the system.

While it is unlikely that AmpliStor or its underlying technology will be widely adopted in the next few years, the simple fact is that increasing capacities of disk drives will eventually make technologies such as AmpliStor a prerequisite in almost any high-capacity enterprise storage system.

So in the same way that enterprise storage vendors started to adopt RAID 6 about five years ago to prevent the loss of data should two SATA drives fail, look for some variation of Amplidata's AmpliStor to find its way into enterprise storage systems over the next decade to prevent the loss of data on these ever larger disk drives. At the same time, expect RAID to find a new home on smaller storage arrays where the level of protection and speed of recovery that RAID currently provides is more than adequate.
This post originally appeared on the DCIG site (see "RAID is Certainly not Dead But its Future Looks Small").