This post is more than 5 years old
105 Posts
0
1671
July 1st, 2016 11:00
RecoverPoint Replica LUN in VNX multi-tier pool?
Hi All,
On a VNX 7600 we are using RecoverPoint to get LUNs from Pool A over to reporting Pool B, then a set of AppSync jobs mounts the LUNs in image access mode (logged) and snaps the LUN, mounts it to a reporting server. Then the reporting server hammers away at the Pool B disks all day.
We keep running out of space (organic growth & VNX snapshots) in the reporting Pool B, that pool is a single tiered pool with lots of 4+4 Raid-10, 900GB disks. The pool is dedicated for RP replicas and only used for mounting to the reporting SQL servers.
I can't find any documentation regarding best practices (consistent performance) to add a NL capacity tier to a pool that hosts RP replica LUNs.
Since RP is always writing the latest changes from the journal into the replica LUN I'm concerned that RP is writing all over the place to random LBA addresses. -IF- that is the case lots of those LBA will be on capacity tier and if a lot of writes come in from PRD before we perform our "nightly refresh" I may be stuck with crap performance on NL all day long.
I also note that any writes from the report server that has mounted the snapshot (Redirect on Write) will also probably be saved in the NL tier.
I have Fast Cache turned on for these replica LUNs, journals are in a dedicated journal pool with fast cache turned off.
Does anybody have any real experience with multi-tiered pools that house RecoverPoint replica LUNs that are used for "reporting" or other OLTP duties? Real world experience with that auto-tiering is going to do would be nice as well.
Thanks,
Ben
forshr
2 Intern
•
1.1K Posts
0
July 5th, 2016 08:00
The write to the replica is the fifth phase of the distribution mechanism and the write profile is random.
Regards,
Rich Forshaw
Consultant Corporate Systems Engineer - RecoverPoint & VPLEX (EMEA)
Data Protection and Availability Solutions
EMC Europe Limited
Mobile: 44 (0) 7730 781169 44%20(0)%207730%20781169>
E-mail: richard.forshaw@emc.com
Twitter: @rw4shaw
forshr
2 Intern
•
1.1K Posts
0
July 4th, 2016 09:00
Hi Ben,
Your concern relating to performance on the RP replicas only becomes an issue, if a highload occurs (async) or if journal lag extends to a point of pausing replication (sync), due to the inability of the throughput bandwidth of the replica pool. The same applies to the journal pool of course as distribution impacts both. Ultimately if you want to guarantee performance then you need to ensure that you have sufficient bandwidth in the pool and/or remove the pool from FAST control.
Regards,
Rich
RecoverPoint Corporate Systems Engineering
benconrad1
105 Posts
0
July 5th, 2016 07:00
Hi Rich,
Can you explain how RecoverPoint writes the LBAs a replica LUN? For example, if production LUN LBA 100 is written and then later re-written with new data what is happening on the replica LUN that is part of the 5 phase distribution? Is the replica LUN LBA 100 going to be written and re-written or is it random LBAs like 200, 300 being written and re-written to (respectively) in the replica?
Ben
benconrad1
105 Posts
0
July 5th, 2016 09:00
That answers my question, re-writes are going to walk all over the LBAs in the replica LUN.
I think it's safe to summarize that if a customer wants consistent performance when mounting LUNs in RecoverPoint Image Access Mode that it's best not to use a capacity tier in the FastVP pool. :-(
Ben