4 Beryllium

Re: Why does RP enter "Regulated" role on source during CDP


In all the release notes there is a section titled Performance with a table listing the "Average Sustained RPA Throughput MB/s". That is where the 75MB/s figure is listed.

It sounds like a pretty frustrating implementation that you are working on. I have been using RP for almost three years now replicating only SQL server databases. We had a lot of challenges intitially with performance and it took a bit of work to finally find a configuration that worked for our SQL CGs. The databases that we are replicating are OLTP type databases with a fairly high iop and change rate. One of the SQL databases is almost 3TB in size. What we found that helped with the replication throughput was splitting the data and log LUNs across multiple CGs and then created a Group Set to keep all the CGs in a consistent synch. I know that you stated that the data LUN is 2TB in size, but is it possible to work with the DBAs to split the data files across multiple LUNs? A SQL performance best practice is to have more smaller LUNs supporting the data and logs to provide increased throughput. This would also enable you to use multiple CGs which would provide more throughput. If this is not possible then you could at least separate the DB and Log LUN in separate CGs and create a Group Set for them. This may help with the init and peak loads.

0 Kudos