I’ve seen front-end storage systems before and never really been too impressed with them. My primary thoughts about a front-end storage system was this, “My storage array already has a front-end, why do I want to put another layer of abstraction in front of it.” Obviously, there is still a use case for having a single namespace to hide the underlying systems, which might be neat so that a company could use multiple arrays of different types or even vendors and have a single place to go to access that storage. For the most part, I still think that this is a luxury that many companies can’t justify since some extra time spent by the infrastructure team will do the job.
Enter Storage Field Day 6 and the presentation by Avere Systems. When Avere did their presentation I was forced to re-think my opinions of a third party front-end storage solution. There were a few specific use cases that I thought were compelling reasons for enterprises to consider a solution like Avere.
The coolest idea was to use their newer Virtual FXT product to provide access to your on-premises storage up to Amazon EC2. Amazon typically has very inconsistent performance when it comes to some of their cloud services, especially S3 storage but it’s very cost effective for a company to do some DevOps on Amazon’s cloud since you can spin down services when a project is done and you stop paying for them. It’s a cloud service, nuff said? Avere allows you to bypass Amazon’s storage system and use your own on-prem storage which is (hopefully) more consistent performance. It does this by utilizing their Virtual FXT appliance that can be spun up on an EC2 image and it will cache reads and writes for EC2 but utilizes your own hardware. I still don’t think this is a fit for everyone, but is an pretty interesting solution.
The second use case is to use these front-end NAS devices for Small Office Home Office (SOHO) to provide much better performance for NFS systems that might be in a branch office, and the primary storage is in the corporate office. This may be a cheaper solution for some organizations than doing some sort of cluster across datacenters or replicating data between filers. Caching hot data in your local site will probably get you “good enough” performance to run a small office environment without needing to buy a large NAS Solution for this small office. You could assume that the Virtual FXT appliance will be expanded to run on vSphere or Hyper-V as well as Amazon EC2. It’s a logical next step, especially since VMware and Microsoft also have their own clouds running on vSphere Hyper-V.
Lastly, we can also assume that we have an aging, legacy filer that is still useful, but not providing the performance that you really need. Throwing an appliance such as the Avere FXT in front of the filer, can handle the read-caching and write-back caching can both improve the readwrite latency with faster drives, hardware etc., but it will also take some stress off of the primary storage array.
Thanks to Avere Systems’ demonstration about their product, I have a newly realized appreciation for this type of service and it really does have a useful story that can be told and it will be added to my utility belt for architectural obstacles in the future.
Check out more information about Avere from the following articles:
All travel expenses and incidentals were paid for by Gestalt IT to attend Storage Field Day 6. This was the only compensation given.