Unsolved
This post is more than 5 years old
1 Rookie
•
30 Posts
0
4073
September 19th, 2014 07:00
Journal sizing in RP
Hi All,
I have few queries related to the Journal volumes. Could any one help with this. Your replies are much appreciated
~ If I have a CG with 5 production luns – each of size say 100 GB (so 5 * 100 GB luns) then while creating the journal, we say it’s a general recommendation to have 20% of the prod lun as the journal volume size. So while creating the journal, do I need to take 5 luns and calculate 20% of it and create journal volume. So 20% of 5 * 100 GB luns = 100 GB journal volume I need to create. Is this correct ?
~ Also for non-Distributed CG, the minimum size of journal lun is 10 GB. Does it mean if a have a journal pool – for eg: 5 volumes then can I create 5 luns * 2 GB also to make journal as a whole to 10 GB ?
~ Also is it recommended to have a one journal volume of 20% of the size of prod lun or is it always recommended to have journal volumes as pool of luns – meaning many luns ?
ankur.patel
2 Intern
•
286 Posts
0
September 19th, 2014 08:00
20% is assuming you want 24hrs of roll back and 20% of your data changes in a 24 hour period
correct 20% of 500GB = 100GB journal
you likely could do 5x2GB, but I cant see a good reason why, there are storage group / view limits which you may hit if you go too small. This feeds into your next question. When you want to grow the 100GB journal you do it by adding another journal lun to that CG. Its best that you expand by adding a lun of the same size as the base so that the data is striped. So in this scenario you would need to add another 100GB lun. Might make more sense to have 2x 50GB so when you expand you add another 50gb.
forshr
2 Intern
•
1.1K Posts
0
September 19th, 2014 10:00
Hi there,
I wanted to add to Ankur's response to your questions/queries.
Firstly, sizing journal space using an arbitrary figure such as 20% of the amoutn of replicated data is a guestimate and should be caveated and reviewed post deployment on the basis of two main factors:
1. Will the 20% satisfy the required protection window, i.e. rollback history?
2. Will the 20% provide enough physical disk drives to sustain peak throughput based on peak MB/s throughput x 3. If it doesn't you may experience slow distribution and/or slow image access.
Where possible I would advocate the use of the EMC's sizing tools to assist you with sizing all the relevant components of a RecoverPoint system and this includes the journalling requirement. However, where perofrmance data cannot be acquired or simulated an arbitrary figure has to be used but I emphasise what I have stated above regarding caveats etc.
In respect of the volumes sizing for a normal CG, you need to ensure that the first consituent volume is 10GB in size. As a best practice I would recommend that you stick to this figure or greater for all constituent volumes.
Regarding having one journal volume for a CG, it is always better to have multiple journal volumes as RecoverPoint performs its own striping across similar sized volume sets, although a pool with striping will offset some of the performance of a single volume allocation. That said, more volumes equals better RPA->Array path allocation and additional queue depth, so as a best practice I would not recommend single volume allocation for a journal.
Regards,
Rich