I know that it seems sometimes like adding filers to the mix may be extra complication, but it really does add a lot of value. There’s good reasons that enterprise shops don’t do local storage.
Basically in ‘real’ businesses, they don’t buy servers with hard drives in them. At all. No drives at all. They do everything on SAN. Even boot from it.
Filers let you de-couple the servers from their storage. This means that if a server crashes (maybe I should say *WHEN* a server crashes), they don’t take the application file systems with them. The files are on the filer. Reload a replacement server, hook it up and go. No recovery from tape required. (If the server is a vm, then you just clone a replacement from a template).
This makes the servers much more lightweight. They don’t have much on them. Just the OS and the app software. The app data lives somewhere else.
Having a filer in the mix makes the vm’s on hypervisors mobile, you can transfer them from one hypervisor to another. If the hypervisors have similar CPUs, you can do this WHILE THEY ARE RUNNING. Try moving a running app from one physical host to another. Obviously you can’t.
The filers can do all sorts of fancy tricks, like scheduled snapshots and replication. So file systems on the filers can be “magically” backed up in the background, without end users even knowing it’s happening.
So it should be apparent that filers are a good thing and that there’s no real reason not to use them.
Except cost.
The most entry level filer at Dell or HP goes for about $5000. Those are plain-Jane ones that do no fancy tricks. Real filers, like Equallogic or Lefthand cost $25,000 and up.
What we have learned to do is reuse older servers and load filer software on them and make them behave like ‘real’ filers. Yes, it may take some fiddling. But with zero hardware cost, and maybe a few hours of fiddling, you get a box that performs *BETTER* than the entry level filers, and arguably approaches the features of the higher end ones like Equallogic.
So even if it takes 10 hours of fiddling at the billing rate (What is that, anyway? Something like $200 an hour?), customers are getting a bargain. $2000 dollars of fiddling to make a $25,000 filer.
So I think the point for customers is that we are SAVING them $5000 or
even $25,000 by using low cost software like Openfiler or NexentaStor. They are getting filer like performance with no hardware cost, and a few hours of fiddling. If they understand that by paying us to fiddle they avoid the $5000 sticker price (or $25,000 price), they won’t balk at a few more billable hours.
Note that what makes the ‘entry level’ filers ‘entry level’ is no snapshot or replication features. You need the $25,000+ filers for that. Both Openfiler and NexentaStor can do it with some (maybe some serious) fiddling. So that’s why I say what we’re doing is better than the entry level.
Of course, real customers with real check books, that properly understand why they need a SAN, won’t want a home grown one, and are willing to pay the big bucks.
Heck, I don’t want a home grown one either. That’s why I have a $50,000 HP EVA 4000 fiber channel SAN, and a $25,000 Equallogic PS4000.