Instance Type for AWS DRS Replication

Question

Hi Team,
I am trying to replicate disks with a total block size of 1.6 TB. Its a multi AZ DR replication using AWS DRS. However, I can see that the Lag and Backlog are very high for initial replication with replication server instance type t3.small, around 8-9 days. I changed the instance type to m5.2xlarge , still the initial replication lag is of around 4 days. I checked the replication server and can see that there is very high CPU overload for t3.small and high CPU overhead even for m5.2xlarge. Is there a specific reason of recommending memory optimized instance types for replication server in DRS instead of compute optimized. I tried changing the instance type to c4.4xlarge and the lag as well as backlog still seems fine of around 2 days. Is the recommendation due to pricing factor or there is other factors involved as well?

Answer

AWS Elastic Disaster Recovery provisions a t3.small server by default. The typical ratio of volumes to replication servers is 15:1.
More info on the replication server instance type can be found here:
https://docs.aws.amazon.com/drs/latest/userguide/replication-server-settings.html

Example:
The replication server instance type can be changed for servers that are replicating too slowly or servers that are constantly busy or experience frequent spikes. These are the most common instance type changes that are made:

-Servers with less than 26 disks – Change the instance type to m5.large. Increase the instance type to m5.xl or higher as needed.
    -Servers with more than 26 disks (or servers in AWS Regions that do not support m5 instance types) – Change the instance type to m4.large. Increase to m4.xlarge or higher, as needed.

Instance Type for AWS DRS Replication

Relevant content