A full filer is not going to perform as well as one that has a little space to do its work. That’s why we’ve been talking about the concept of a “thin provisioned” filer, or a filer that has been “thinned out” by taking out older, unused data – leaving room for the filer to breathe. Let’s take a look at what NetApp has to say about this issue. Is it true? Do full NetApp filers perform worse than thin provisioned ones?
Before answering this question, we need to discuss the terms volumes and aggregates. An aggregate is a pool of storage into which you create a volume. Traditional NetApp volumes had a one-to-one relationship with their aggregate; however, there can be multiple FlexVols within a single aggregate.
NetApp has the concept of reallocation, which is basically defragmentation of data on a NetApp. It is disabled by default, but can be enabled and scheduled to run on a regular basis. The reallocation process needs room to work, and free space is therefore essential for it to work correctly. There are different levels of reallocation runs, and some levels may need more room to work than others. If there is not enough free space, a reallocation process may do a whole lot of work, only to determine there is not enough free space for it to do what it needs to do. When this happens, the reallocation process exits with an error.
If reallocation cannot run on a volume or aggregate, you have to run it on a file-by-file basis, which is significantly more difficult and time-consuming than the regular process. It’s also not going to increase performance as much as a full reallocation run.
This is why NetApp recommends keeping at least 10 percent free space in a volume or aggregate. The technical note Configuring and Tuning NetApp Storage Systems for High-Performance Random-Access Workloads mentions that in addition to giving optimization room to work, leaving at least 10 percent free space “also functions as a buffer against unexpected demands of free space for applications that burst writes to disk.”
Also remember that snapshots need room to work as well. Snapshot copy reserve sets aside a percentage of the disk space for storing snapshot copies. If that space is no longer available, snapshot copies will spill into the active file system in a process called Snapshot spill. This will further exacerbate the low disk space conditions mentioned above.
This is why we’ve discussed this idea of moving infrequently accessed data off of the filer and onto less expensive object storage where it will be automatically protected. If you can do that in such a way that you can still access the data in its original location – all the better.
NetApp volumes and aggregates need room to breathe. A NetApp filer will perform better if the reallocation process is run on a regular basis, and it needs free space to effectively reallocate blocks and put like data together on the disk. Consider thin provisioning your NetApp so that it can increase its performance.
Sponsored by Caringo