ORIGINALLY POSTED 30th October 2014
25,721 views on developerworks
With half term week here in the UK and preparations for next months travels, I almost forgot that yesterday we announced the contents of version 7.4.0 of the SVC and Storwize software that will be GA’d on 28th November.
The full RFA can be viewed here, but here’s a quick summary.
Upgrading to 7.4 will enable (on supported platforms) two instances of the RACE (Random Access Compression Engine) software – which in short means almost twice the performance for most workloads. The supported platforms are the new SVC DH8 hardware – when the node is capable of compression (2nd CPU, extra cache and hardware offload card(s)) – and the Storwize V7000 gen2, again when all the compression assist hardware is installed (extra cache and both compression offload cards).
The allocation of volumes is essentially round robin across both instances, so you do need at least two volumes to make use of this enhancement, but it takes random IOPs performance from say 175K up to over 330K when the workload behaves well with compression. The bandwidth has also increased significantly and almost 4.5GB/s can be achieved per SVC node pair.
Encryption for Data at Rest
The 12G SAS hardware that makes up the disk backend for the Storwize V7000 gen2 has been capable of hardware (zero performance overhead) encryption since GA in June. With 7.4 software installed you can now create new arrays that encrypt the data as it flows to disk. This means you do not require SED (Self Encrypting Drives) and can encrypt data on any of the supported SAS HDD or SSD drives. The feature is licensed per system and once enabled all new arrays created on the system will be encrypted by default. Each array has its own key by default. Initially local key management is provided, and the master keys are stored on USB sticks at the time of array creation. This encryption support means drives themselves can be protect from malicious removal – and if drives fail and need to be replaced you can be sure that nobody can read the data from the failed drives. A statement of direction to bring data at reset encyption for Fibre Channel storage virtualized under SVC and Storwize and full support for industry standard key management has also been made.
16Gbit Fibre Channel HIC
Fairly self explanatory! You can now order new SVC DH8 or Storwize V7000 gen2 systems with dual port 16Gbit FC interface cards, Existing systems can be MES upgraded to 16Gbit. Each DH8 SVC node can contain up to 4 of these dual port cards, giving 16x 16Gbit ports per IO group (I’ve measured almost 26GB/s read hit from such a system!) Each V7000 gen2 can have 2 of these cards per node canister, hence 8x 16Gbit ports per IO group.
Increased Global Mirror round trip latency
Up until now we have supported up to 80ms round trip latency for GM links. While at the speed of light thats 8000km in theory, nobody has perfect links and they rarely travel in straight lines! In 7.4 we now support up to 250ms round trip latency and distances of up to 20,000km are supported. Combined with the performance improvements in 7.2 software, these changes and enhancements have greatly improved the reliability and performance even over poor links.
T10 DIF Support
7.4 introduces the industry standard extension at the RAID and SAS level to provide an extra level of data integrity. The use of this type two protection information (PI) between the internal RAID layer and SAS drives appends 8 bytes of integrity metadata while the data is being transferred between the controller and the PI-formatted disk drives. The 8 byte integrity field contains CRC data and more that provides validation data that can be used to ensure that data written is valid and is not altered in the SAS network. For more information see http://www.t10.org
Feedback from end users regarding drive replacement procedures told us that in certain scenarious you didn’t want to have to follow the DMP to complete to ‘exchange’ of the old for new. Now, you can simply swap the drives, and the new drive in that slot will take over from the failed drive.
There are often cases where you want to sub-divide a storage pool (or managed disk group) but maintain a larger number of mdisks in that pool. Child pools are logically similar to storage pools, but allow you to specify one or more sub divided child pools. Quotas and warnings can be set independently per child pool.
Increased FlashCopy consistency groups
You can now configure 255 FlashCopy consistency groups – up from 127 previously.
Protect volumes on delete
This optional feature stops administrators from inadvertently deleting an active volume! Yes it does happen… You can configure the “idle time” from 15 to 1440 minutes. That is, the volume has to have been idle for that long before you are allowed to delete it. You can of course force the deletion by disabling the feature if you want.
VLAN support for iSCSI
You can now use VLAN tagging within iSCSI host access.
Support for 19 expansions on V5000
As the title suggests, from 7.4 onwards you can now configure a V5000 control enclosure with 19 expansions, and cluster two of these together to support 960 drives per system.
I think thats about it… another busy year coming to closuring with this 7.4 release, and lots more already being worked on for 7.5 next year!