For over a decade, network attached storage systems (NAS) have tried to expand beyond the typical use case of a store for user home directories. Companies like NetApp maintain NFS is perfectly appropriate for supporting virtual infrastructures and Oracle databases. The performance advantages of a NAS system vs. traditional array are questionable, but “ease of use” is not. With the advent of all-flash filers, performance is becoming a non-issue for many data centers. That said, an all-flash filer has to deal with the reality that all data does not have to be on flash at all times.
Thanks to an all-flash option, a NetApp Fabric-Attached Storage (FAS) has the potential to be the single storage repository for all production data. However, at some point these systems will run out of capacity. IT will then face the decision of expanding the current system or buying an additional one. There is also the performance issues of a “full filer” that my colleague Curtis Preston discussed in his article “Performance Issues with a Full NetApp Filer”.
All-flash filers make these issues even worse. First, because of their performance and built-in data efficiency, organizations are far more likely to consolidate down to a single system. This means the filer will reach its capacity limits sooner than a hard disk drive system. Also, because of its high performance, IT professionals may think that it is acceptable to run these filers closer to their maximum capacity. But the issues of a full filer still exist, even if that filer is an all-flash system. In fact, to some extent, flash exacerbates the problem and its low latency makes performance bottlenecks even more conspicuous.
Finally, there is the obvious cost disadvantage of investing in more and more flash. Despite the gains in cost per GB and the addition of data efficiency technology, flash is still more expensive than the equivalent hard disk capacity. Using this premium storage for data that does not explicitly need its performance is a waste of IT budget.
Data that is either inactive or does not need the performance of an all-flash storage system should move to less-expensive, probably hard disk-based, storage. A thin-provisioned filer performs and manages data protection tasks better.
FAS Perfect for Object
Interestingly, unlike traditional block storage, a NetApp FAS makes the movement of data to an object storage solution relatively easy. Since most object storage systems support the NFS protocol, movement between the two platforms can be a simple copy or move command. But as we discuss in our webinar What Your Object Storage Vendor Isn’t Telling You About NFS Support, not all object storage systems support NFS in the same way. Understanding the differences between types of NFS support is critical, especially with an all-flash FAS.
NetApp, as my colleague Joseph Ortiz discusses in his article “Getting Data Off a Filer and Living to Tell About It” creates a very useful set of APIs that enable vendors to cleanly interface with the FAS file system (Data ONTAP). Armed with support of these APIs, an object storage vendor could routinely and automatically migrate inactive data off the filer and onto their object store. More importantly, they can set up links to the new file location so that when users attempt to access that old data, they can do so transparently.
An all-flash FAS has the potential to be the single storage system for an entire organization as long as its scope is limited to production data. An object storage system has the potential to be the single storage system for all other data, including inactive data on the filer as well as analytics and machine-generated data. The combination leads to a cost-effective, easy-to-manage, two-system environment with a clear delineation of each system’s role.
Sponsored by Caringo