Hello,

With SVC 8.5.0.10 there is:

SVAPAR-107595 Improve maximum throughput for Global Mirror, Metro Mirror and Hyperswap by providing more inter-node messaging resources.

Can this also have a positive impact on stretched cluster?

Thanks.

------------------------------
TMasteen
------------------------------

Hello,

If you are using gm, mm or hyperswap just run the "test only" option with the latest test utility for 8.5.0.10. 

------------------------------
Nezih Boyacioglu
------------------------------

Hiya - 

Just to clarify a bit more on this APAR:  This change in behaviour is to address a drop in performance seen when going from SV1 to SV3 nodes using Remote copy specifically.

It aims to fan out the work a bit better as the increase to 48 cores in the SV3 caused resource contention.  Remote copy wasn't really designed to scale out this way and performance decayed instead of maintaining or increasing as a result.  This change has some pleasant side effects on high core count systems, but it is most noticeable on SV3 which was the aim as SV1 clients need to be moving to SA2 or SV3 systems.  This is not a problem for the new Policy based replication code which is architected to scale and perform and is what people running 8.6.0 should be looking to migrate to in the medium term.

Stretched Cluster relies on Volume mirroring technology, not remote copy, so unless you are ALSO using remote copy in your stretched cluster you won't see an impact.

I hope that helps.

------------------------------
Evelyn Perez
IBM Senior Technical Staff Member
IBM Storage Virtualize Software Architect for SVC and FlashSystem
------------------------------

Totally helps, thanks for your clear answer.

------------------------------
TMasteen
------------------------------

Original Message:
Sent: Thu December 14, 2023 08:46 AM
From: Evelyn Perez
Subject: Q: SVAPAR-107595

Hiya - 

Just to clarify a bit more on this APAR:  This change in behaviour is to address a drop in performance seen when going from SV1 to SV3 nodes using Remote copy specifically.

It aims to fan out the work a bit better as the increase to 48 cores in the SV3 caused resource contention.  Remote copy wasn't really designed to scale out this way and performance decayed instead of maintaining or increasing as a result.  This change has some pleasant side effects on high core count systems, but it is most noticeable on SV3 which was the aim as SV1 clients need to be moving to SA2 or SV3 systems.  This is not a problem for the new Policy based replication code which is architected to scale and perform and is what people running 8.6.0 should be looking to migrate to in the medium term.

Stretched Cluster relies on Volume mirroring technology, not remote copy, so unless you are ALSO using remote copy in your stretched cluster you won't see an impact.

I hope that helps.

------------------------------
Evelyn Perez
IBM Senior Technical Staff Member
IBM Storage Virtualize Software Architect for SVC and FlashSystem
------------------------------