RE: Thoughts about storage

[Rich Graves <rgraves@carleton.edu>]

> What are folks monitoring on the OS side to signal storage performance issues?
> IOSTAT svctime, atime, avgqu-sz?

zmstat-chart might already be gathering what you want. Take a closer look. If you're really hitting hardware limits, though, the ontap layer is the place to focus, and you can poll ntap's private mib with mrtg/cricket/whatever.

On the software side, zipping blobs (with or without compression) can be evaluated as a way to reduce metadata ops, and hence decrease seek/increase sequential write ratio, though that would likely blow away any ontap-layer dedupe advantages. See wiki/forums.

While I wouldn't do it, you could consider just not running zmbackup, relying on snapshots mountable as VMs instead, assuming that netapp's overhead for read/write snapshots is reasonable. I dunno; I last drove a netapp > 4 years ago. You can still achieve point-in-time recovery by keeping redologs, just as you would with oracle. There's a good discusssion in the bugzilla that added zmplayredo in 5.0.7 (or thereabouts).