Why the CLARiiON doesn't require 100% snap reserve ...

Chuck's analysis of usable capacity was interesting. He took best practices, divorced those practices from customer requirements, and cherry picked configurations to highlight the CLARiiON's usable capacity. His biggest attack on NetApp storage centered around the 100% snapshot reserve requirement NetApp FAS systems used to have.

Rather than explain why NetApp had a 100% snap reserve,I think it's far more interesting to ask the question why the CLARiiON never did.

CLARiiON snapshots because of the performance crippling implementation as well as their space inefficiency are by their very nature temporary transient objects. In many cases, a CLARiiON snapshot is really just a disk-stage for tape and not a backup image that would be used for recovery.

Heck, according to EMC's deleted white paper on CLARiiON snapshots, the expectation was that if you care about  performance on an Exchange system you wouldn't  even use snapshots (Never mind 10% reserve, they recommend 0% reserve and not to use the feature). And if even snapshots are used, the expectation is that the snapshot will be kept around for as little time as possible. In fact, the primary use case for CLARiiON snapshots is not to do recovery, but to provide a disk-de-stage for tape backup. .

So why does the CLARiiON have a 10-20% reserve?  Because a snapshot that is deleted once every twenty four hours, will not consume very much space.

On the contrary, for NetApp storage systems we expected, and our customers expected to use those snapshots to do recovery from a day, a month or a year ago. And they can do that because our snapshots impose no  performance penalty and use the minimal amount of space.

NetApp Snapshots form the basis of NetApp's very successful Exchange backup and recovery tool SnapManager for Exchange.

So why 100% reserve? Well if a snapshot is going to be kept for a month or a year, it is not unreasonable for a storage admin to assume that every disk block will change. Now remember, NetApp snapshots are backup images. If you over-write a disk-block, we have to keep track of the original disk-block to be able to perform a recovery.

When we started out building SAN systems, we took a very conservative position with respect to snapshots and space.

Over time, as we got more sophisticated with our  understanding of our customers and as our system advanced we realized we could offer more options. Not everyone wanted to keep snapshots for very long periods of time. Not everyone wanted to keep them on all on FC disks. Some people wanted to keep snapshots around for a long while as long there was space, relying on SATA disks or tape for archival etc…

The point is that the reason  the CLARiiON never had a best practice requiring 100% snap-reserve is because no storage admin who ever cared about Exchange performance kept a snapshot around for any protracted period of time. The need to keep 100% over-write reserve, never actually occurs on a CLARiiON system because of the limitation of the CLARiiON.

So if you want a less capable device, and you want to recover from a remote device, go ahead use a CLARiiON. Good news is that if you want that and want the flexibility to keep more data on the primary to have faster recoveries, we can do that too…

Comments

[Geert]

On deleted white papers, did you notice Chuck's remark at the end of his blog entry:

From http://chucksblog.typepad.com/chucks_blog/2008/08/your-storage-mi.html:

"Warning -- if some of these docs disappear after we post this, our apologies"

Ain't that ironic....?

[Kostadis Roussos]

You know, that was the most disingeneous, aspect of his post.

[Dan]

Better still - with volume auto grow and snapshot auto delete options available to the Netapp administrator, it's entirely feasible to set the guarantee type to Volume, drop factional reserve to 0% and allow the volume to consume space from the aggregate as and when it's needed. Stick that in your pipe and smoke it CLARiiON.

Hard drive recovery saves files from being lost indefinitely. Software is the most efficient solution so if your drive is working try it first. A data recovery agent can be contacted for in-lab service.