HP released today an update for its MSA storage line. So what, you may think? Well it is a quite important one.
This update will add several virtualization features into the entry-level storage MSA portfolio. And there is also a really cool new GUI interface.
By installing this new firmware you will get 6 new functions:
For those who have known the HP EVA storage, they should know that this was the first array that used storage virtualization. Independent of the RAID type selected, all your volumes were spread across all disks in the array, delivering the highest IOPS due to the fact that all spindles were delivering I/O.
In the old ‘legacy’ way of working the only thing we had to create was a vDisk which was a container of a certain amount of physical disks in a specific RAID set. The maximum amount of disks was 16 in 1 vDisk so customers with a lot of disks had to create multiple vDisks. Knowing that a volume cannot span a vDisk this was the limiting factor for the size and the performance of that volume.
In the new code a new functionality being Virtual Storage Pools are added. These pools will be used to ‘glue’ several disk groups together. Knowing that a volume will be created in the pool and not in the disk group means that the volume will have the performance of all disks in all disk groups combined in a virtual pool.
When the pool is full with data, very easily a new disk group can be added online, all data will be automatically distributed over all disks in the tier. Capacity and performance will go up instantly.
By leveraging the virtualization layer in new MSA FW (GL200 or newer), the system allows a user to present a Volume(s) which are much larger than the physical storage resources actually allocated to the volume. Physical resources from a thinly provisioned volume will be allocated as they are used.
Users can deploy large virtual volumes to support applications without having to purchase and allocate the physical storage to that application on day one. This allows a “pay as you grow” deployment model. Thin provisioning virtually eliminates the need to expand volumes later caused when traditional linear volumes grow initial volume size. You also dramatically cut disk rebuild times via “thin rebuild”.
Important to know is the T10-Unmap (space reclamation) support from day 1, something we cannot say (unfortunately) for the StoreVirtual storage line. Promised since a long time, still not there… This functionality is very important to be able to use in an efficient way thin provisioning.
SSD Read Cache
The new code allows to create a new type of Disk Group specific for SSD Read Cache. The 4GB cache in the controller can be extended with additional SSD drives in this Disk Group, which will enhance the read performance by copying frequently accessed (random) data to SSD. Sequential data will remain on HDDs to optimize performance/value.
In this example you see the IOPS on the Standard Tier diminish as data is copied to the Read Cache.
The Tiering functionality will use the new virtual storage pools (see wide striping) so that the blocks of data (pages of 4MB) will be placed on the most ideal type of disk, based on their I/O access type. Random I/O will be most of the time on SSD, sequential I/O will be mainly on SAS and infrequently accessed data will be on MDL SAS (7,2K rpm) disks.
This Tiering is permanently active and automated, unlike Adaptive Optimization on 3PAR that needs to be scheduled and so configured.
Tiering between MDL SAS and SAS is descrived as Archive Tiering and is free on MSA 2040. On the MSA 1040 this can be actived by adding an supplemental virtualization license.
Tiering to SSD is also called Performance Tieringg and is only available on MSA 2040 (since 1040 does not support SSD drives). This performance tiering requires an additional license on the MSA 2040 as well.
Advanced Virtualized Snapshots
The original snapshotting mechanism used was a Copy-On-Write action. Not bad as such but not very efficient because of the write penalty.
By using Redirect-On-Write now less I/O’s will be needed giving a lower performance impact.
By using this new mechanism allows also to create snapshots of snapshots, something that was not possible in the old code.
The biggest ‘visible’ change is the new SMU (Storage Management Utility) GUI. Really nothing of the original is maintained. On one view you get an instant overview on the health and performance of the device and all assigned storage pools, disk groups and volumes. By hovering the mouse over the several sections of the console you will get instant pop-ups with additional information. Very efficient.
In those few weeks I was able to use and test the new code I must admit that it all works really fluid, fast and stable. I have other experiences with the previous SMU consoles in the past. Not this time. HP learnt from the past apparently. Well done!
All these new features will be available by installing the new GL200 firmware.
The MSA 2040 will have all features enabled for free except the performance tiering to SSD. For this you need to purchase an additional license. On the MSA 1040 the virtualization features can be enabled by adding a specific virtualization license. Since the MSA 1040 does not support SSD, only archive tiering will be available, and also SSD read caching is not available due to the same reason of no SSD support.
The update procedure is straight forward like all the previous updates
The MSA family exist of 2 models: MSA 1040 and MSA 2040.
MSA 1040 is positioned for customers where budget matters. Lower ports per controller (2), fewer controller options, max 99 disks, no SSD support.
MSA 2040 is proposed when performance matters. 16Gb FC support, more ports per controller (4), converged controllers for more flexibility and efficiency, SSD support, 199 disks supported with 7 additional disk enclosures.
HPStorageGuy aka Calvin Zito created a cool ChalkTalk on this announcement as well:
1 Remark: Where is the StoreWhatever name for MSA? 😉
Be social and share!