What do you get?
VDI workloads tend to DeDupe well, rates in excess of 10:1 is not uncommon.
For general purpose VMware workloads the benefit is unfortunately not as stellar. On average we tend to see around 0.5:1 for DeDupe.
Compression tends to provide 2:1 data reduction so combined with DeDupe a good rule of thumb is to expect 2.5:1 from your new Storwize code.
Look Before You Leap
So before you all go scrambling to IBM Fix Central to download the new DeDupe code (V8.1.2) here are few things to think about.
- Existing pools needs to be migrated in order to active DeDupe. You will need to create a clone volume and turn on DeDupe for that volume. Once the volume copy is complete then shift the workload to the new volume.
- Not all workloads DeDepe well. For example, databases don't often benefit from DeDupe. Be selective about where you turn on DeDupe. If you'd like to assess your environment for DeDupe friendly workloads you can run the handy IBM Data Reduction tool. (Search for 'DR Reduction Tool' in Fix Central)
- What is your code level methodology? Typically I recommend at least N-1. Bear in mind that DeDupe is currently in the latest version of Storwize code. I'd wait for at least one new increment version to come out before installing on your productions machines.
The DeDupe code available takes advantage of the existing hardware offload cards that we already use for compression. This allows you to implement DeDupe with minimal impact on the CPU.
Devices with hardware offload: V7000 Gen 2, V7000 Gen 2+, SVC DH8 nodes and SVC SV1
Devices with CPU only data reduction: V7000 Gen 1 and V5030 (Use data reduction with cation)
If you're thinking about implementing DeDupe on your existing IBM array to get some free storage myself and the team are here to help. Feel free to reach out to us or your friendly IBM Business Partner to discuss.