In a VMWare dual site (one cluster) setup, which is the "smarter" storage to use for datastores? I kinda like SMAC's ability to direct read/writes to the right storage that is closest to it. As far as I know, this has not yet been implemented in MCCIP, where you just have a normal SVM on one of the sites, and you can end up creating a VM in site A which is on a datastore local for site B... (We are talking iSCSI or NVMEoTCP)... Are there some new MCCIP features on the way that I haven't yet discovered on this front?
#SnapMirror Active Sync vs. MCCIP in a VMWare dual site setup
floral quartz
crisp sedge
you are right that in a block-only world, SMas is the nicer solution as it lets you direct the writes to the system that's closer to the server. However, cross-datacenter latency still applies and writes still have to go through both sites before they are committed, so there's no real performance gain in such a configuration. The only advantage is that those writes go through the (potentially faster) backend network (peering) instead of the frontend (iscsi) network
however, MCC IP is still the smarter solution if you want a "fire-and-forget" solution, or you have NAS workloads that benefit from site protection
floral quartz
Yes I understand, but in a MCCIP don't you still have to keep in mind on which datastore you place your VMs ? You could potentially end up with cross traffic? Or can you setup iSCSI paths on both clusters from the same SVM?
crisp sedge
yeah, definitely, you would get lots of front-end traffic if you place the VM on the wrong datastore on an MCC IP
floral quartz
I think I remember that in the "older" MC's with FibreChannel frontend you could access data from both nodes? But maybe this was all the way back in the 7-mode days? 😉
But can you setup rules in vmware that control where the storage is placed when you create a new VM? Or do you have to be awake at that point? 😉
crisp sedge
I think I remember that in the "older" MC's with FibreChannel frontend you could...
yeah, but those were still just a "regular" HA pair (single image) that just had twice the disks in the right location to survive a desaster. Since MCC IP are two separate clusters, it's not so easy in the cDOT world anymore 😦
crisp sedge
But can you setup rules in vmware that control where the storage is placed when ...
not sure about that, but in any case, you can (manually or automagically?) SVmotion the VMs if you notice they're placed incorrectly
floral quartz
Would be nice if NetApp could "borrow" some of SMAS's features... Maybe it's in the works...