Today’s the first day of the
Green Grid Technical Forum conference (Compellent is a member), so it's appropriate to think about how companies can reduce data center costs while also reducing the carbon footprint of their data centers. Through some new ways of thinking about storage you can reduce the overall number of disks needed and reduce physical space needed. That might sound odd coming from a co-founder of a SAN company – after all I'm rewarded for the amount of storage systems our company sells – but reducing your storage usage is the best, long-term approach to the problems many businesses are facing now.
So what’s the recipe for an energy efficient SAN? Here are five key ingredients:
- Thin provisioning – It’s a must have for a cost-effective, “green” data center. Remember, by increasing the amount of utilization – the percentage of the drive that’s taken up by actual data, not just allocated space – you’re going to lower the number of overall drives required. You’re no longer pre-allocating storage capacity to volumes before that capacity is used.
- Automated tiered storage – Distributing data across a combination of disks by continually tracking usage and automatically moving data between tiers based on predefined rules and frequency of access. By automatically moving infrequently used data from tier 1 storage, which is almost always power-hogging 15K RPM Fibre Channel drives to energy-efficient, higher-capacity SATA storage, you can greatly reduce the number of high-energy drives. And as a result, cut down on the cost of powering your storage system.
- Storage virtualization – You need to merge all your disks into a single pool of storage, allowing any volume to use all disk drives simultaneously to access data. That way, you can separate applications from physical storage devices to get the same performance with fewer drives compared to a non-virtualized storage environment.
- Boot from SAN – By putting an OS boot volume on a SAN rather than on the server, you don’t need internal server disks. You can rack up diskless blade servers to save space, power, cooling, and cabling costs. You could save thousands of dollars per server by going diskless, too. But beware the fine print. Not all boot-from-SAN solutions are equal. Make sure that you don’t need third party agents or apps (which means internal server disks!) to point your servers to the "gold image" stored on the SAN.
- Continuous snapshots – Using a multi-tiered storage system lets you store read-only data such as snapshots on cheap disks. Some snapshot technologies are more space-efficient than others – with pointers to changed data instead of mirrored data. That way, when a snapshot is taken, the storage within that volume becomes read-only, so that it doesn’t need to be written again. That means instead of RAID 10, you can use RAID 5 and save yourself some drives, and just as important, use less power and cooling than the RAID 10 configuration for active, read-write data.
Like I said
yesterday real estate is a prime commodity and many businesses don’t have space to spare for larger data centers. If your IT initiative for 2008 is to reduce the number of disk drives required – and therefore reduce the amount of space and energy you’re consuming – take a look around at the innovative technology available today. I don’t believe we need to compromise data storage functionality for energy efficiency, any more than we’d compromise our favorite family recipe -- we're just updating it. Let me know if you agree there is such a thing as an energy efficient SAN. In a future post, I'll analyze the positive impact on disk drive and energy usage using the above storage technologies in a data center.