Netbackup, Puredisk and Compression- no one knows...
I’m trying to learn as much as I can about PD but even Symantec own documents and www pages contradict themselves. I’ve got to many examples to list but here’s a little gem I found last night. That page is also linked to by other “how to” docs… sad.
I still have questions no one seems to be able to answer.
· Ok, so they say to enable compression for PD using the media servers pb.conf file.. but what about clients using “client side dedupe”? The only way I can find is to edit each client’s pd.conf file.
· Also, new versions of PD 6.6.1+ use a new type of compression but the OST plugin for PD on the NBU servers does not get updated when you update PD so what does this mean about what kind of compression media servers/clients are using? Release notes for PD updates basically say “if running NBU 7 on Win/Linux check back for media server updates for PD plugin”
· It’s not talked about and hard to find info about compression at the 3 levels where dedupe can take place, client, media server, PD servers. It seem like Symantec hasn’t thought much about it ether. Because as far as I can see the 3 places are all treated separate and you have to choose where you enable it and manage on a per client/server basis.
· Documents warn that after enabling compression the new compressed data will be deduped different from uncompressed data so you’ll see dedupe rates drop for a bit until uncompressed data expires. So what does this mean in light of media servers running a version of PD plugin, clients running a version of PD plugin, and the PD servers running a possible different version of PD and compression? Wouldn’t this also have quite a negative affect if compression is enabled on the media servers but you don’t edit each and every client using client side dedupe to also enable compression as they will be sending uncompressed data to the PD pool?
When I ask Symantec techs these questions there is only silence or guesses on the other end of the line.
In light of what seems to be the need to edit each client if using client side dedupe and having no idea what type of compression method the media servers are using, it seems the best idea is the enable compression on the PD servers themselves. This way I know it’s using the most current version of compression PD supports, I don’t have to wonder if client X has compression enabled. Although this would add CPU load to the PD servers I’m seeing almost no CPU usage on them currently. Adding compression to client side dedupe also adds cpu load to servers I don’t directly manage. I’d rather monitor load on a hand full of PD servers than hundreds of clients. Clients and media servers would process hashes and the PD servers would do the compression of incoming data.