SVC and Storwize V7000 Real time Compression

Last month IBM released the 6.4 code for SVC and V7000. The main feature this code release introduced was Real time Compression (RtC). Being over at Storage Bootcamp in IBM Montpellier last week I got the chance to perform some testing on RtC, global mirror with change tracking on SVC & V7000 and creating automated migration policies for file systems on V7000U (Blog posts on global mirror with change tracking and automated migration policies for file systems will be posted up soon)

The impressive IBM Montpellier complex

In 2010 IBM purchased Storwize an Israeli outfit for a figure reported to in the region of $140 Million. Storwize produced appliances that sit infront of NAS arrays and compress data being written to the arrays via Lempel-Ziv algorithms in its Random Access Compression Engine. (RACE) The RACE engine is the only storage solution that can compress active primary data without effecting performance and typically achieves a data reduction ratio of upwards of 50%. 

Example of embedded RACE compression

The increasing rate of data growth is a growing concern for almost every organisation. Therefore the ability to compress primary data without impacting performance potentially presents a major cost saving for any storage subsystem that can be virtualized under the SVC or V7000 can benefit from. A compression estimator utility, the Comprestimator has been made available via IBM to estimate the expected compression ratio for block devices (The Windows version of the Comprestimator is currently in beta). The Comprestimator estimates the portion of non zero blocks in each volume and the compression rate of non zero blocks. From this you can gauge which volumes are good candidates for compression. Workloads estimating 45% or more are ideal candidates, 25%-45% are potential candidates, however the workloads will always need reviewing. Typically collabaration data, databases, e-mail, engineering data and server virtualization are the most suitable workloads for compression. Approximate expected compression rates for specific data types are shown in the table below.

Collaboration Data	30% - 75%
Databases	50% - 80%
E-Mail	30% - 60%
Engineering Data	50% - 80%
Server Virtualization	45% - 70%

So you have upgraded your SVC or V7000 to 6.4 and have noticed the changes to the GUI but how do you convert your current volumes to compressed volumes? Unfortunately there is currently no dynamic conversion process. To perform conversion process you need to follow this process:

1)   Create a mirrored copy of the volume you wish to compress 

2)   Specify the mirrored copy volume type as compressed 

3)   Allow the volume synchronization to complete 

4)   Promote the compressed volume (copy1) to be the primary volume 

5)   Confirm the impact upon CPU utilization via the Compression % within the CPU graph and once confirmed the volume is performing as expected

6)   Delete the old non compressed volume 

The conversion process therefore needs sufficient swing space for the compressed volume mirror copies to be created. To estimate the amount of space required run the comprestimator, review the output and confirm if you have enough space within your storage infrastructure to perform the conversion process. As always there are a few gotchas to watch out, I have listed these below:

• Compression requires I/O groups hardware to be on one of the following platforms:
• V7000 2076-1xx/3xx Control Enclosure
• SVC Model 2145-CF8/CG8 Nodes
  *6.4 also listed as supported on SVC Model 8A4, 8G4, 8F4 & 8F2 Nodes 
• Maximum of 200 compressed volumes per I/O group
• CPU recommendations for 4 Core Systems (V7000, CF8 & older CG8 SVC nodes) to be utilising less than 25% of CPU before created compressed volumes
• CPU recommendations for 6 Core Systems (newer CG8 SVC nodes) to be utilising less than 50% of CPU before created compressed volumes
  *If existing system CPU utilisation is over the thresholds specified above in environments with less than 4 I/O groups a new I/O group can be added to support compressed volumes within the cluster
• The CPU reallocation is completed as soon as the first compressed volume is defined
• Compressed volumes are currently not supported via EasyTier
• Compressed volumes only currently supported on block level storage (compression on file level storage support expected within next release of SONAS)

Some further useful information on RtC can be found within the links below:

ESG Lab Report on V7000 with RtC
IBM Storwize V7000 Software 6.4 Introduces Real time Compression
Real-time Compression in SAN Volume Controller and Storwize V7000

From a licensing perspective compression is licensed per Terabyte on the SVC and per enclosure on the V7000. IBM are offering a free 45 day trial of RtC, details of this program can be found here

If you are interested in discussing the benefits of Real time Compression and how these can be introduced within your environment or have any queries regarding the blog post above please feel free to comment or get in contact via wbush@virtualvizion.com