Group Member Striping Algorithm

[crw64]

I am trying to understand how the member load balancing algorithm works in a group with multiple members. I mean, I understand that connection load is striped across the arrays, taking advantage of the increased cpu, cache and network, and that the volumes get intelligently stripped across the group members. But when does this striping occur? For example I have a member with a number of volumes. Adding a new member array to this group, I notice that no load is actually distributed across the newly added array. However once I create a *new* volume in the group (once both arrays are now members) the second newly added array is taking connections for this volume, like the original member. IE. The new volume was created across both arrays.

So my understanding is that existing volumes will intelligently start striping out across the newly added group members. But how long does this take to begin? And is the volume evenly split 50/50? I know that with our external disk shelves, the volumes will begin filling the disk shelves until they are equally populated at which point they will start striping data across both simultaneously.

[crw64 · Accepted solution]

Nevermind. I found the information I was looking for in the Scale Out Best Practices guide.