Further information on How traffic is split between replica nodes

Software-based VM-centric and flash-friendly VM storage + free version

Moderators: art (staff), anton (staff), Anatoly (staff), Max (staff)

Post Reply
bkdilse
Posts: 52
Joined: Mon Sep 10, 2018 6:14 am

Thu Oct 04, 2018 11:20 am

Hi,

Is there further information on how traffic is split between the a 2 node Storage only setup (no Compute).

I have a 2 node setup, using Synchronous “Two-Way” Replication, and wondered how my iSCSI VM traffic on ESXi/Hyper-V Compute servers is distributed across the 2 nodes.

I did read about this, but cannot find the article again.
Oleg(staff)
Staff
Posts: 568
Joined: Fri Nov 24, 2017 7:52 am

Fri Oct 05, 2018 11:27 am

You can check this article.
bkdilse
Posts: 52
Joined: Mon Sep 10, 2018 6:14 am

Sun Oct 14, 2018 7:22 pm

thanks, but that wasn't the one I read. I'm sure I read something on the lines of, IO is split between nodes depending on load/IO already hitting a node.
Oleg(staff)
Staff
Posts: 568
Joined: Fri Nov 24, 2017 7:52 am

Mon Oct 15, 2018 9:54 am

You are speaking about MPIO policy, am I right?
bkdilse
Posts: 52
Joined: Mon Sep 10, 2018 6:14 am

Mon Oct 15, 2018 10:39 am

No, MPIO part is fine. What I'm interested in is, how the Starwind Service decides which of the 2 nodes receive the IO requests.

e.g. 4 VM's with VHDX hosted on the Storage Only VSAN. Does the Disk IO for the VMs hit radom VSAN Nodes, depending on which is busy, and do they keep splitting between nodes, or do they stick to a paticular node?
Oleg(staff)
Staff
Posts: 568
Joined: Fri Nov 24, 2017 7:52 am

Mon Oct 15, 2018 11:52 am

I got it.
This information you can find by following this link:
"StarWind HA performs in active-active mode, thus it is a best practice to use identical hardware for all the nodes participating in an HA storage cluster. Rarely, one HA node can have faster storage to improve read performance. In this case, ALUA (Asymmetric Logical Unit Access) is configured to achieve the optimal performance. ALUA marks a certain storage path as optimal or non-optimal for write IO. The initiator server (if supported) uses these marks to optimize LU access. Please refer to the Asymmetric configurations chapter of the document for more information. "
bkdilse
Posts: 52
Joined: Mon Sep 10, 2018 6:14 am

Mon Oct 15, 2018 12:05 pm

Thanks Oleg, that's the paragragh I was after :)

Just noticed, that paragraph in the other link you sent, I must have browsed past it. Thanks again.
Oleg(staff)
Staff
Posts: 568
Joined: Fri Nov 24, 2017 7:52 am

Mon Oct 15, 2018 12:21 pm

You are welcome :)
Post Reply