HP P4000 Snapshot clarification

[A.W.R]

Hi,

We have an HP P4300G2 starter SAN.

We use snapshots fairly frequently.

In the documentation it states that the snapshots are :-

Snapshots provide a copy of a volume for use with backup and other applications. You create

snapshots from a volume on the cluster.

Snapshots are always thin provisioned.

Am I correct in saying that over time the snapshot will grow in size as new data gets written to the volume, and the orginal data gets written to the snapshot? So if I have multiple old snapshots this can be quite an IO penalty?

Is there a performance penalty for deliing a snapshot?

Your feedback on this would be really appreciated.

Thanks

Andrew 

[Dirk Trilsbeek]

the snapshot systems i know usually store the new data in the snapshot, not the old data. So the snapshot grows over time until it is removed and all the commits within the snapshots are written to the volume. The main performance hit should be the fact that each read first has to check if there is a new version of the data in the snapshot(s) before reading unchanged data from the volume.

[Bart_Heungens]

Hi,

The P4000 does not work like the 'other' snapshots though the result is the same...

P4000 will not copy out the original block to somewhere else and then write the new block inside the volume...

P4000 will keep the volume like it is and will write the new block in a seperate location behind the original volume... When the OS wants to read data it will read from the top down to the bottom, thru all the eventual snapshots above... If it finds a valid block in the highest shell, it will take the block and continue... If it is not in the highest shell, it will go one below and so on until it finds that block...

There is a good picture in the training material that explains this, but I cannot share this on a public forum since it is an official HP curriculum...

Kr,

Bart