I’m interested in the new feature allowing ESXi hosts using a “Round-Robin” policy for multi-pathing to VPLEX instead of fixed-path.
This has been presented during "VPLEX advanced configuration" during last EMCworld 2014.
Since "fixed-path" config was the "best-practise" for EMC, where can I find some more info about the new feature? (VPLEX version, ESXs version, benefits, pros and cons, etc.)
Internally we're working to update our documentation around recommending round-robin over fixed path. It'll be probably a couple of weeks before it's published publicly. The EMC VMware host connectivity guide will be updated.
The main benefit of recommending round-robin over fixed path, is that for fixed path it was very manual and difficult to get the lun ownership balanced properly across all VPLEX directors. For hundreds of hosts and thousands of volumes you can imagine how difficult of a task this was. Not to mention that even if you did happen to perfectly manually load balance things, then on a path failover from the host it is completely arbitrary which path was chosen as the next path.
Round-robin makes all this much easier, and behaves in a similar manner to any round-robin multipathing policy which is what PowerPath Adaptive (intelligent round-robin policy based upon path latency and outstanding I/O's) has been doing from the very beginning. There is the potential for some additional inter-director cache-coherency traffic however that additional traffic is very minimal and far outweighs the potential for imbalance that existed with Fixed path policy.
VMware's Compatibility Guide will outline which ESXi versions go with which GeoSynchrony versions.
I'd like to see the EMC docs updated too. But if you want something official, the VMware Compatibility Guide does list Round Robin. So I'm going to go ahead and recommend RR, except in the case of cross-connected Metro Cluster.
It would be nice if we could define a preferred director or preferred ports for each volume in a storage view, and have VPLEX advertise those ports as ALUA optimized. That would be a way to eliminate the manual load balancing on the host side, and the issue with inter-director cache traffic.
garyo
89 Posts
3334
1
Posted May 22nd, 2014 23:00
Hi,
Internally we're working to update our documentation around recommending round-robin over fixed path. It'll be probably a couple of weeks before it's published publicly. The EMC VMware host connectivity guide will be updated.
The main benefit of recommending round-robin over fixed path, is that for fixed path it was very manual and difficult to get the lun ownership balanced properly across all VPLEX directors. For hundreds of hosts and thousands of volumes you can imagine how difficult of a task this was. Not to mention that even if you did happen to perfectly manually load balance things, then on a path failover from the host it is completely arbitrary which path was chosen as the next path.
Round-robin makes all this much easier, and behaves in a similar manner to any round-robin multipathing policy which is what PowerPath Adaptive (intelligent round-robin policy based upon path latency and outstanding I/O's) has been doing from the very beginning. There is the potential for some additional inter-director cache-coherency traffic however that additional traffic is very minimal and far outweighs the potential for imbalance that existed with Fixed path policy.
VMware's Compatibility Guide will outline which ESXi versions go with which GeoSynchrony versions.
VMware Compatibility Guide: Storage/SAN Search
In short, GeoSynchrony 5.2 and 5.3 will be round-robin preferred along with newer ESXi versions (5.1, 5.5.)
Hope that helps,
Gary