I’ve done some testing internally, and at this time I don’t see a lot of benefit when spinning disk is involved. It definitely compresses, but as drives get
bigger and bigger customers tend to be fewer of them. Then end result is the compression and decompression activity might push you over the edge into performance problems. It’s generally a better option to just use an extra spindle or two rather than trying
to save it with compression.
There are exceptions, but we’d really need to know the specifics of the environment. For example, if you have a 32k database block size you might even see a
performance increase from compression because that’s a nice large block to work with. If the database is mostly inert, then you have little to fear in general from putting extra load on the system and it seems to work well. Pure-flash environments don’t have
the performance hit from a physical head flying around the spindle, so compression is less of a performance drain too. You’ll never get something for nothing, though.
The best route is frequently to use database compression at the database layer. It’s a big mistake to enable that everywhere, but I’ve seen a lot of customers
enable it on certain tables to improve performance. It might save space too, but that’s not the point. Let’s say you have 10:1 compression. If you start reading a table on a 8Gb FC port, you’ll be able to read 80Gb of logical data because you’re moving it
around compressed. That’s why you’ll see compression more and more with things like the new in-memory options, HCC, HANA, Hadoop, and other things. If you keep data compressed, you can store more of it in RAM and move it around more efficiently. Again, there
can be a performance impact if not used wisely.
Finally, I would encourage anyone interested in compression to contact their account team for further information on compression options. If they have the specifics
of what you’re like to do, we can share additional details.
From: toasters-bounces@teaparty.net [mailto:toasters-bounces@teaparty.net]
On Behalf Of Peta Thames
Sent: Tuesday, June 10, 2014 7:58 PM
To: Toasters@teaparty.net
Subject: Oracle Volumes and Compression
Hi,
I've read that compression works well on volumes hosting Oracle DB data, but I haven't seen it in use anywhere in the wild yet. Is anyone using compression, and if so, what have been your experiences?
Cheers,
Peta