Getting Started with NetApp Storage Efficiency

This is my advice for customers who want to get started with storage efficiency:
  • Consider SATA drives instead of Fiber Channel
  • Enable Dedupe
  • (Use Flash Cache as insurance against bad performance)
NetApp has other efficiency features too (thin provisioning, cloning, compression, and so on), but I’ve found that customers often start with SATA and dedupe. SATA because it saves so much money, and dedupe because it’s so easy to turn on and comes free with ONTAP. When I talk with customers who are using SATA and dedupe, they are usually happy with NetApp, and pleased with their storage costs. When customers are haggling over price but haven’t at least considered these features, I wonder what they are thinking. SATA with Flash Cache doesn’t always match the performance of Fiber Channel, but when it does, it can cut your costs in half. It’s definitely worth considering! We have many happy customers using it for production data. Home directories are a good place to start. Email, especially with the most recent versions of Exchange. Some customers use it for database, depending on the workload. The nice thing about dedupe is that it’s so easy to turn on and off. It’s best to turn it on when you first create a new volume, because then data gets deduped as it arrives. It might seem like a good idea to wait until the volume is full, and you need the extra space, but that doesn’t work as well. First of all, it takes CPU and disk bandwidth to dedupe all that data at once. And second, what is in snapshots is read-only and can’t be deduped, so you have to wait for old snapshots to age out. Better to turn it on from the start. But if you didn’t do it then, do it now. (Depending on workload, dedupe can impact performance. If it does, turn it off. That’s easy!) The usage of dedupe varies widely in different parts of the world. For instance, in Denmark, over 50% of all disks that we have installed have dedupe enabled. (This is based on autosupport data.) Wisconsin and Missouri are both at 45%. Is there something special about Danish data that makes it more dedupable? I don’t think so. Rather, I think people in some areas have just gotten in the habit of turning it on. And it works well, so they tell their friends also to turn it on. (When I look only at systems installed in the past 12 months, Denmark is over 65%.) Friends don’t let friends drive NetApp without dedupe.