A consistency group (CG) is a logical entity that defines a collection of virtual machines and their disks that are to be replicated together in a way that ensures write-order consistency.
A single RecoverPoint for Virtual Machines system can replicate up to 512 CGs.
A CG runs on a single vRPA and the maximum performance (IOPS and throughput) of that vRPA limits that CG.
When several CGs run on the same vRPA, they share the resources of that vRPA. Higher-priority CGs consume more resources than lower-priority CGs.
For copies, the replica copy journal configuration affects the replication performance of a CG. For best practices in configuring these journals, see Choosing a datastore for journals.
When a new CG is created during VM protection, this new CG is assigned to one of the vRPAs in the vRPA cluster in a way that balances the protected capacity per vRPA. Typically, not all vRPAs have the same I/O load. Hence, you should manually modify the assignment of CGs to vRPAs to balance the I/O load. For additional information, see the RecoverPoint for Virtual Machines HTML5 Plugin Administrator’s Guide or RecoverPoint for Virtual Machines Flex Plugin Administrator’s Guide.
Most RecoverPoint for Virtual Machines replication operations such as enabling image access for a specific point in time, and failover are done per CG. Hence, they are performed on all the VMs that the CG protects.
If you need a single-VM granularity for these operations, then you should put each VM in a dedicated CG.
Before doing so, however, you should consider the following scale and performance factors:
The VMs that are replicated in a CG are always consistent. In some situations, however, due to performance considerations, it is not always possible to replicate the VMs in a single CG using a single vRPA. In this case, you can protect these VMs using several CGs, and put all these CGs in a group set with parallel bookmarking enabled. Group sets are easy to manage and guarantee consistency across the CGs that belong to them for every parallel bookmark.
A group set with parallel bookmarking enabled has the following limitations:
I have 4 VMs that depend on each other. Each generates approximately the same I/O load, and together they generate 50,000 write IOPS. How should I protect them with RecoverPoint for Virtual Machines?
50,000 IOPS is greater than the maximum IOPS supported by a single vRPA. Hence, you must divide the load over several CGs in a group set with parallel bookmarking enabled. According to Table 2, one option is to protect two VMs per CG. In this way, each pair of CGs runs on its own "8‑vCPUs, 16‑GB RAM" vRPA, and each vRPA sustains 25,000 IOPS.