Windows file servers are primarily used to store unstructured data. The type of unstructured data varies from user-created documents to machine-generated audio and video. They even store analytics data machines and other servers generate. While the type of data a Windows file server may store varies, all this data has one thing in common, when created it is initially very active then is rarely accessed again. We design these servers for the initial rapid access use case but expect them to fill the role of long term repository.
Windows File Servers As a Cache
It’s time to change the role of our Windows file servers. Or better stated, start designing them for the role they were intended for. Windows file servers should essentially be a cache (or tier if you prefer). They should be fast, have plenty of network IO and almost no capacity. That’s right, almost no capacity. The capacity they do have should be flash-based SSDs. With this design they can be even more responsive to the initial use case, providing fast response to creation and modification requests. The moment, or soon after, that data is no longer modified or accessed it should be removed from the file server.
File Server Cache Backend
When that data is removed from the file server cache it obviously has to move somewhere and for most organizations the ideal “somewhere” might just be an object storage system. Object storage is designed for long-term preservation of data. It is very cost effective, highly redundant and scalable. It is also responsive to recall requests, which, if the file server is used as a cache will occur frequently. The speed of recall is critical, it has to happen fast enough that users won’t notice a difference in recall speeds or if they do, it won’t be different enough that they will complain about it.
Traditional caching software works because it operates seamlessly to the users and the applications. For a Windows file server to act as a cache to an object storage system, the “caching” software has to transparently move data between the two. Unlike a traditional caching solution the movement of data in this design is not only between two different types of storage systems (NTFS vs. Object), it more than likely is a movement of data between different vendors.
Ideally just like a traditional cache, this software will copy data as it changes between the two storage types. If so that means like a cache, IT does not have to worry about backing it up. The object store is the initial backup.
Adopting a Windows file server caching mentality allows IT to design a dramatically less expensive file server while offering significantly better performance. And, putting a cap on Windows file server sprawl. Then the object store can do what it does best, preserve information for an indefinite period of time.
Sponsored by Caringo