This post is more than 5 years old
1 Rookie
•
85 Posts
0
4842
January 21st, 2013 05:00
RecoverPoint journal configuration in VNX environment
Currently designing a RecoverPoint/CRR solution and seeking for some good best practices.
The setup consists of 2x VNX7500 with FastCache and RecoverPoint 3.5SP1, in a campus environement (<5km distance, plenty FC bandwith between sites). RP is configured to use FC for replication traffic.
RecoverPoint is basically used to provide replication for DR purposes, image access mode will be used only for testing purposes.
The VNX systems consists of two large FAST autotiering 3-tier-pools per VNX, 1 pool for Unix and 1 pool for VMware. All luns will be replicated and 50% of the primary (source) luns will be on VNX1 and 50% on VNX2. In RP two Consistency Groups will be made. 1 for Unix and 1 for VMware. Failover to the other VNX for only a single server or application is not expected.
The point is how to size the journals and where to place. My own experience is that best performance can be achieved by using a dedicated pool for journal luns. In this case we expect to need 10TB of journal size, 5TB per CG. We want to use 25x 600GB SAS drives for this purpose.
Question 1: does it make sense to go for 30GB journal lun size here? With 1,5TB per CG this would result in 50 journal luns per CG.
Question 2: is there a risk that RP will stripe across those luns (which will be in the same pool with the same underlying disks) resulting in degraded performance of this pool?
etaljic81
1K Posts
0
January 21st, 2013 06:00
It really depends on how much performance you need from those Journal LUNs. Comparing RGs vs. pools; you can see up to a 10-15% performance benefit if you go with RGs. If you want the best performance for the Journal LUNs go with RGs.
Go with small LUNs for two reasons: it's easier to expand Journal capacity and RP stripes the Journal data across the LUNs. If we are talking about 4 RGs dedicated for Journals it doesn't make sense creating more than 4 Journal LUNs for a CG (4x250GB) unless you are looking to expand the Journal capacity for 100-150GB later on.
etaljic81
1K Posts
0
January 21st, 2013 05:00
I recommend you seperate the journal drives from the two pools you mentioned. You state: "We want to use 25x 600GB SAS drives for this purpose." Is your plan to configure all of those 25 drives in one pool and use that for Journals? Or, did you plan to use regular RAID Groups?
Creating 30GB LUNs for Journals doesn't make sense. Why did you decide to do this route? RP will stripe across the Journal LUNs but if those 30GB LUNs are from the same pool it doesn't make sense. If you need a 1TB journal for a CG I would recommend to create 2x500GB LUNs for the Journal which resides on the 25-disk pool (assuming you are going with a pool design). Why 500GB and not 1TB? This way you can expand the Journal capacity later on by adding another 500GB LUN instead of a 1TB LUN. RP BP is to keep the Journal LUNs the same size.
jpveen
1 Rookie
•
85 Posts
0
January 21st, 2013 06:00
Ok, thanks for sharing your opinion!
jpveen
1 Rookie
•
85 Posts
0
January 21st, 2013 06:00
Thanks for your quick reply. Plan is to use those 25 drives for a dedicated pool for RP-journal luns.
Traditional raidgroups are not planned, to provide easier administration and upgrades in the future. Do you think that traditional RG's will have significant benefits over a pool in this situation?
About the lunsize: RP trainingdocu for example advices to go for those small luns. But unclear to me either what the reason is to split up the journal for a CG to more than let's say 2 luns.
etaljic81
1K Posts
0
January 21st, 2013 06:00
No problem. Glad I can help.