Best Practice for RP Consistency Groups and Data Warehouse components?

Question

We use RP currently to replicate our main set of databases to a remote site (as well as to CDP). The consistency group is comprised of five different luns.

Now I am prototyping SAN to SAN replication for an upcoming EDW project. The components to be replicated will include not only the same OLTP databases we are currently replicating, but also the staging, CIF (data warehouse), dimensions, data marts and analysis cubes.

My question is: what is the recommended setup for consistency groups? Would it be better to put all of these into one consistency group? Or should I split some out into a different group? I worry that if they are all in one group then pausing replication for one part means pausing for all. I also worry about sizing the journal if they are all together. However, if I separate them, then this can have potential impacts for restoring to a single point in time, right?

What is your recommendation?

JamesBEMC · Answer

Hi there

RP is absolutely great with EDW's, as RP cares only about the disk writes, the disk reads will pass through, and EDWs typically have high read:write ratios.

My recommendation is to do the following;

* Use RecoverPoint 3.3 so that you can use the grid-CG functionality to harness the throughput of all available RPAs.

* Keep your OLTP data in the same consistency group and set a highest priority in the WAN utilization policy (OLTP data is more important than DW data)

* Create multiple CGs, so multiple Journal volumes

1 - your EDW's staging (ETL area), (normal priority)

2 - CIF (higher, not highest priority)

3 - Dims, marts and cubes (normal priority)

Then create a group set to run at whatever desired frequency in order to drive cross-CG consistency across your entire EDW.

That means you can recover your entire EDW from multiple CGs in a crash-consistent manner.

Questions:

* Bandwidth between sites?

* I assume your staging (pre or post ETL?) area expericences heavy burst writes?

Thanks

James

sleverson · Answer

Very very helpful, thank you. Our bandwidth between sites right now is 20 Mb.

Our Staging will indeed experience heavy write bursts (actually, so does our OLTP due to batch processes that occur nightly). We will have two or more ETL processes in place at one time.

JamesBEMC · Answer

Good stuff.

My recommendation would be to use the Grid-CG option in RecoverPoint 3.3 so.

Also, keep an eye on the environment in terms of replication for the first week or so to see if the policies need tuning.

I'm sure your local EMC architect will help here.

Best of luck. Any more Q's dont hesitate !

James.

RecoverPoint

Best Practice for RP Consistency Groups and Data Warehouse components?

Was this post helpful?