Compression

  • 28 February 2014
  • 8 replies
  • 7878 views

Userlevel 2
Badge +6
Is anyone using compression? If so, for which workloads and what is the experience like on your cluster?

8 replies

Userlevel 4
Badge +21
I have a customer who's using it with great success for databases.
Dedup wasn't getting the job done (the same was true when he was on NetApp).
Compression helped things out, with no noticeable performance impact.

Sylvain.
Userlevel 2
Badge +6
That's helpful, thanks. What kinds of databases? What size were they? What kind of space savings are you seeing?

As far as I know, dedupliction is only on the hot tier in NOS 3.5.x. So we won't see any savings by employing deduplication until Nutanix releases NOS 4.0.
Userlevel 4
Badge +21
Databases were Oracle 11g, as I recall, size was in the 200-400Gb range.
I don't know about the space savings, I'll have to check with them, but they were happy enough.

I'll get back to you if/when I get the chance to see them.

Sylvain.
Badge +5
In Nutanix documentation we could see such slide
Is it really usefull to use inline compression for DB?

Badge +5
I wouldn't use inline compression for databases. At most I'd use post process compression with a long delay time. Great for the type of databases where you're just storing stuff for a long time and not writing to it again.
Badge +3
Hi,

have been running Nutanix for a couple of months now and have successfully used inline compression for our SQL DB servers.
I have moved about fifteen 0.5 to 1 TB big VM's over from a very competent NetApp enviornment and I have seen no performance impact of compression. I use PAGE compression within SQL server and I have still been able to get 1.8 : 1 in compression ration.

I have made extensive tests both with and without compression and I have seen no negative effect of the inline compression so far. General load of the cluster is still low as I have a 8 node 6060 cluster, my only concern is what happens when CPU load on the cluster increases. Remains to be seen.
Badge +5
Hi there.

I've been trying to understand compression. Looking on NutanixBible, I can read this about inline compression.
No impact to random I/O, helps increase storage tier utilization. Benefits large or sequential I/O performance by reducing data to replicate and read from disk.

However, when looking at the , it seems that inline compression for random I/Os has an impact. 1st of all, it needs to coalesc data before to compress. Therefore, in case of read of any chunk into the coalesced bloc, it will have to decompress all of it. So I guess this is true for any write, and any change in records of a database right ?

In that case, Inline compression should only be used for dump databases where data are written once/never read ?! Are there any comparison benchmarks about real (and not feel) performance impact ? I can't say to a customer "use compression, it doesn't feel worse" ... 😉
Badge +2
Hello,

Can anyone please post your recent experiences and/or recommendations regarding Compression and Deduplication? We're setting up our first VMware vSphere environment on Nutanix (AOS 5.5) that will run mainly database workloads (Oracle and MSSQL) and a few application servers. I've read quite a few postings and Nutanix documentations on this topic but there seems to be a lot of contradictions/disagreements so I'm still confused as to which direction to go. Thank you.

Reply