Thin provisioning a NetApp can create as many problems as it solves. As mentioned in previous blogs, thin provisioning in this context refers to the act of clearing out older, unused files and putting them somewhere else. Essentially “thinning” the capacity of the filer to only include the most active data. This has a number of advantages, including increasing performance, data protection, and availability for your NetApp filer.
But if you’re going to move old data out, the question is where are you going to put it? We discuss how to get it back in our blog “Getting Data Off a Filer and Living to Tell About It”. Three choices come to mind: the cloud, another filer, or an object storage system. They each have advantages and disadvantages.
Moving data to the cloud is an obvious choice for many environments. The main advantage is it does not require a large capital outlay to begin the project. You simply need some type of application to identify the older files and migrate them into S3, Glacier, or similar service. Unfortunately, this advantage can also become a disadvantage.
Customers moving a lot of data into the cloud can find themselves with a very large monthly cloud bill, as well as the latency and cost issues associated with a WAN network. Customers finding themselves in the situation often discover it is easy to get data into cloud vendors – not so easy to get it out. For example, most cloud services have a tape in service that allows you to ship them a box of tapes that they will import into their system. Good luck finding a cloud service with a tape out function. In addition, in order to stop paying the monthly charge, you must transfer the data out of the cloud – incurring the large per gigabyte cost of such a transfer. So while the cloud may be a great choice for some; others can find the recurring costs quite daunting.
Another choice is to purchase a different kind of filer that is better at handling large data sets. There are a number of scale-out filers on the market, but you have to question the wisdom of moving data that has a very low value from one high cost filer to another. If you do move the data from one filer to another – even if the other filer is a scale-out system with significantly more capacity – the data will still need to be backed up with the production backup system. One of the reasons for moving the data off the filer in the first place is to exclude it from backups, and placing it on another NAS system that needs to be backed up seems like an obvious case of kicking the can down the road.
An object storage system, on the other hand, has the scale-out advantages of a scale-out filer, but at a significantly reduced cost. In fact, the acquisition and recurring costs of an object storage system are significantly less than NAS filers. In addition, since object storage systems are self protecting, one can safely move older data into an object storage system and stop backing it up with the production backup system. Finally, while there will be ongoing operational costs for an object storage system, they pale in comparison to the monthly costs from a cloud vendor that has terabytes of your data.
Customers with smaller amounts of data (measured in dozens of terabytes) are probably best served with cloud storage. If they ever find their recurring costs unsatisfactory, they could migrate to one of the two other options. Moving the data to another filer seems fundamentally wrong and the worst of the three choices. However, if a customer has hundreds of terabytes to petabytes of older data that needs to be stored as cheaply as possible and retrieved quickly if necessary, an object storage system would seem to be the best bet.
Sponsored by Caringo