The highlight of the IBM EDGE 2012 is the announcement of compression on the IBM Storwize array. Two years ago IBM acquired Storwize, a file compression appliance company. IBM applied the name to a mid-range unified storage array (Storwize V7000, based on SAN Volume Control (SVC) technology) and renamed the appliance as the IBM RTC (real-time compression) appliance. Two years later, IBM has announced a block-based storage version of the technology on the same V7000 array. The compression is performed inline in real-time by the active-active storage controllers.
IBM RTC Technology
The compression technology uses the standard Lempel-Ziv (LZ) algorithm. The Storwize patents acquired by IBM refer to the ability to apply LZ in real time and for random data. The recovery of data stored on disk is assured by a LZ-based utility supplied by IBM. LZ compression is lossless and is used in a wide variety of data processing implementations, including tape drives. The average compression ratio is in the range 2-3, with lower figures for media files and higher figures for text-based and general database files. For the V7000, Wikibon recommends a net figure of 2.0 (50%) saving be used.
File-based storage on the V7000 is implemented as file-on-block, using a modified GPFS file system. For compression, it still requires the IBM RTC (real-time compression) appliance for the moment, but Wikibon is confident that file-based storage will be added to the controller version in the near future.
The RTC implementation on the V7000 is new, and time will be needed for a final assessment of RTC on the V7000. Yoni Cohen, CEO of Snowball Studios, presented an interesting proof-point at EDGE 2012. In his environment, the average project size was reduced from 417GB to 148GB, with complete transparency to the user and no impact on performance.
Wikibon CORE Methodology
In 2010, Wikibon introduced CORE (Capacity Optimization Ratio Effectiveness), a methodology for measuring the effectiveness of storage optimization technologies as a function of the time and cost to achieve a desired capacity reduction. The original Storewize technology scored very high indeed on the CORE scale (171, where 1 is break even), the highest of all compression technologies measured.
Wikibon has not yet formally reviewed the detailed implementation of RTC on the V7000 nor formally assessed the CORE ratings for RTC on the V7000. Wikibon expects that the figures would be lower that the current RTC ratios, but still very good, and best-of-breed for block-based array compression.
Block-level compression has become standard on midrange arrays, including the EMC VNX and NetApp arrays and IBM n-series arrays. The CORE rating for the NetApp and IBM n-series are currently assessed as 0.4 and would only be recommended in very low IOPS environments. The EMC VNX has a similar implementation to the NetApp compression, and Wikibon expects that the EMC VNX would have similar CORE ratings. However, the Wikibon CORE ratings have not been formally updated for the latest NetApp and EMC arrays.
IBM Storwize V7000 assessment
Storage arrays have two components, the drives themselves and the controller technology. The addition of RTC to the IBM V7000 will improve the utilization of the disks drives but will not significantly reduce the number of IOs or controller costs. Wikibon agrees with IBM that the net saving is about 25-30% for typical V7000 environments. The savings would be lower for higher IOPS configurations.
Wikibon places the IBM Storwize V7000 as best-of-breed compared with the IBM midrange storage arrays, including the IBM XIV and the IBM n-Series. The reason for this are the combination of capabilities:
- The ability to efficiently combine file and block-based data on a single array, with a common and highly effective GUI interface.
- The availability of space efficient technologies such as thin provisioning and space efficient snapshots.
- The ability to extend the life of existing array assets by virtualizing them and managing them as an extension of the V7000 (and reducing the cost of software on those arrays).
- The availability of automatic tiered storage (IBM Easy Tier) on the V7000 (not available on the n-series or XIV), which maximizes the benefit of SSD disks and reduces the overall cost of storage. This provides a more flexible environment than the SSD read-only caches on the XIV and n-Series and applicability to a broader set of workloads.
- The quality of the additional controller functionality inherited from the SVC controller (e.g., replication).
- Last and by no means least, the availability of real-time inline compression.
Other IBM mid-range arrays do offer specific functionality that could benefit users in specific environments (e.g., de-duplication on the n-Series, greater caching on the XIV), but overall Wikibon would place the IBM Storwize V7000 as the most cost-effective and versatile IBM midrange storage array.
Action Item: CXOs and senior storage managers should include the IBM Storwize 7000 for all midrange RFPs.