Archive for August, 2011

VNXe Deduplication and Compression

August 22, 2011

I had to do some digging recently to get some technical details on the deduplication technology imbedded in the EMC VNXe arrays so I thought I’d share it.

Summary

The file-level dedup on the new VNX arrays is only for Shared Folders and NFS datastores.  This deduplication is policy driven and utilizes both file deduplication (single instancing) and compression.  The compression of files has NO impact on data sharing.  The deduplication acts on whole files and does not look at the metadata.  In addition, shared folders and NFS datastores can have different dedup settings and the dedup will work in conjunction with replication, snapshots and file level retention.

Policy Engine

The policy engine which controls deduplication runs automatically and is governed by high and low CPU watermarks for the Datamovers.  When the CPU is below 40% the policy engine runs at full speed.  When CPU utilization is between the low and high watermarks (40 – 75%) the policy engine is throttled.  CPU utilization above 75% will cause the policy engine to go into a paused state until the utilization drops.

There are 2 main settings that can be configured in the policy engine, the File Extension Exclude list and the Path Exclude list.  Here you can set certain file types or data paths to be excluded from deduplication. 

Additional Thresholds

The threshold for deduplication on files is 30 Days.  Files that have been read or modified in the last 30 days are exluded from the deduplication process.  There is also a file size threshold to be aware of.  Files greater than 8TB and smaller than 24KB are excluded from the deduplication process also.

So, while you certainly won’t see deduplication ratios like you would on DataDomain or Avamar appliances the VNX deduplication can prove to be very useful and save you precious production storage space.  Best of all, VNX deduplication is easy to turn on and manage through the Unisphere GUI. 

Advertisements

vSphere 5 vRAM Entitlements Get Larger

August 7, 2011

To say the least it’s been an interesting couple of weeks in the world of VMware.  First – the big vSphere 5 announcement – Yay!  Then the realization that VMware screwed around with the licensing scheme again and added vRAM entitlements – not so Yay!  In fact, at least from the clients we’ve spoke with, most of them were straight up pissed off, and understandably so.  The screaming was so loud from the customer base that around a week later the details of the new licensing were, well, let’s say “re-evaluated”…… yeah that sounds nice.  Essentially for the Enterprise and Enterprise Plus products the vRAM entitlements were doubled and for the Standard Edition vRAM was raised from 24GB to 32GB per processor.  Truthfully this should eliminate any issues for most customers but the real question is how much damage has been done with this new vRAM brain child?

In a recent WebEx the point was made that with the new revision to vRAM entitlements something like less that 3% of existing customers will be affected.  That’s great but if that’s the case why create this new model in the first place and get everyone worked up?  On top of getting everyone worked up VMware also got many people thinking about alternatives such as HyperV and Citrix XenServer.  That can’t be a good thing.  In fact, I was hanging out with a buddy of mine just this weekend who works for Microsoft.   In his words, Microsoft is absolutely jumping for joy over this latest licensing announcement and they should be.

In my opinion VMware is still the best virtualization platform out there.  They make a great quality product and if this new version of vSphere makes them more profitable in a competitive market place -more power to them.  It just seems to me with companies like Microsoft and Citrix gaining market share and looking for ammunition – VMware just left the door wide open.  Let’s face it, price isn’t everything but in the current economy more and more people are paying closer attention to costs.