20 December 2017

What Makes SVC So Special? - Technical Benefits of SV1 Nodes and the Non-Disruptive Upgrade Process


Taking a trip down memory lane I have been fortunate enough to work with all of the IBM SAN Volume Controller (SVC) models from the 2145-4F2 that was released way back in 2003 up to the 2145-SV1 which was released late in 2016.

The thing in my mind that makes SVC so special is how it enables simplified management of Storage infrastructure. 

Enhancing storage functions via enabling key technologies such as the Virtualization of external storage, real-time compression, automated optimised tiered storage via EasyTier, replication, stretch clustering, encryption and enabling the movement of data between dissimilar storage subsystems without disruptions. 
These technologies are primarily enabled via the Spectrum Virtualize software running on the SVC nodes. 
IBM are further enhancing the capabilities of Spectrum Virtualize on a quarterly basis, they now offer a software only solution and also now offer Spectrum Virtualize as a Cloud based service in IBM Cloud enabling hybrid deployments of Spectrum Virtualize. The great thing about this is it enables DRaaS quickly and efficiently.  
There's many more exciting developments coming in the near future with SVC and Spectrum Virtualize that I will discuss in later blog posts.

Running SVC means your Storage environment is seamless to manage and not dependant on the physical Storage hardware. The great news is this means you are not locked into the technologies delivered via single vendor, customers in this scenario without SVC often end up with silos of Storage and an environment that is very labour intensive to manage. 
I have also seen SVC be a live saver in terms of enabling organisations to cope with entire site outages with minimal impact to end users. 
This obviously needs the correct Storage Area Network setup at the backend to enable a stretch cluster to be deployed but if designed, configured and deployed correctly it is extremely effective.
In this day and age where data is the lifeblood of organisations it really in my opinion is the best Storage Management engine in the world. 


The latest generation of SVC the SV1 delivers increased performance and additional internal storage capacity. 
Each SV1 node consists of two Xeon E5 v4 Series eight-core processors and 64GB of memory with the ability to upgrade this to 256GB. 
It includes 10 Gb Ethernet ports standard for 10 Gb iSCSI connectivity and service technician use, and supports up to four I/O adapter cards for 16 Gb FC and 10 Gb iSCSI/FCoE connectivity. It also includes two integrated AC power supplies and battery units. Those who remember the CG8 and older nodes will remember the standalone UPS's, these are no longer required by the DH8 and SV1 nodes due to the 2 integrated battery units per node. No lugging UPS's into the rack and reduced power bills which is a bonus! 

As previously documented in my blogpost around Spectrum Virtualize 8.1 prior to the 8.1 release the SVC memory manager (PLMM) could only address 64GB of memory per node. In 8.1 the PLMM has been rewritten, structure sizes have been increased enabling more memory to be addressed via the cache if available. There are no visible changes apart from larger read cache sizes and therefore improved performance. 
The SVC SV1 and V9000 have had the capability of supporting 256GB per node hardware wise since v7.8.1 but the v8 operating system allows the cache to fully take advantage of all this lovely memory to improve performance and ultimately improve the end-users experience. Therefore if you have SV1 nodes and want to have more than 64GB per node addressable you need to be running Spectrum Virtualize 8.1 or later. 

I had the pleasure of completing a CG8 to SV1 migration a few weeks back. Needless to say this isn't a simple procedure but if done correctly can be done fully non-disruptively. 
The key is to ensure the replace SV1 node uses the same worldwide node name (WWNN) as the node that is replacing. My main advice is ensure you document the WWNNs and get the FC port map translation confirmed prior to starting the migration. 
The full end to end process is documented via IBM here   
Typing rmnode is a horrible proposition as effectively your first step is to remove the node from the system before replacing with the new SV1 node. But once over that hurdle and providing you have the completed the required pre-reqs, have the required knowledge and understanding of the process then all should run smoothly. I managed to migrate 4 nodes from CG8s to SV1s within a day. 

It becomes apparent during the migration process how much quicker the SV1 nodes are than the old CG8s. 
Full node reboots happen in under 10 minutes compared to the previous timescales of 45 mins+. This means migrations to these nodes are much quicker and also firmware upgrades are much quicker.  

If you need any help planning SVC upgrades or want to discuss the most effective way to enable Storage Virtualization within your environment feel free to get in contact with me via william.bush@tectrade.com

No comments:

Post a Comment