How to set up a single data aggregate (ONTAP 9.1P5) but avoid I/O going through interconnect.
2018-03-16 07:44 AM
Question from a customer setting up new FAS8200s with 3-4 shelves of 1.8TB SAS drives:
Config: set up two aggregates, one to each node, each with ½ the disks, and a single SVM for each data aggr and node. If the disks are “Shared”, does that mean both SVMs can write directly to the underlying disks without going through the interconnect links?
Customer is trying to avoid having any I/O go through the interconnect link, and wants to know if shared disks makes this possible. Or if having junction paths “connecting” the two SVMs makes it possible.
Ideally, they would like to have a single large data aggr that can be accessed equally by both SVMs via both nodes, without going through the interconnect link, but don't konw if that is possible. They can't use FlexGroups at this time.
Another way of asking this:
“Does node ‘ownership’ of the disk determine the use of interconnect [for I/O] , or does SVM ‘ownership’ define it?”
1 REPLY 1
Re: How to set up a single data aggregate (ONTAP 9.1P5) but avoid I/O going through interconnect.
2018-03-18 11:52 AM
Aggregates are tied to nodes. Aggregates cannot span nodes. FlexVol volumes live on single aggregates. You can set a data LIF on each node and have clients only access volumes based on their locality.
But avoiding cluster interconnect traffic isn't really necessary in most cases, and in fact, complicates things overall.