EMC’s Recoverpoint has always fascinated me. The technology that manages to split writes out and use those to journal these writes and replicate them is awesome technology. Unfortunately as with many cool technology features, it was complex and prone to error if not doen exactly right. Today EMC announced something that makes very cool technology WAY MORE cool: Recoverpoint will sit inside the hypervisor. What you can do then is mind blowing!
A very quick intro into EMC recoverpoint
Recoverpoint is basically a snapshot and replication technology that is independent of the used storage architecture. And much more. And with limitations. Basically recoverpoint uses appliances, which used to be physical appliances. these appliances would capture all writes on a SAN, and store them in a journal. Capturing the writes is done using a “write splitter”: These write splitters manage to see all writes passing through, and duplicating them to the recoverpoint appliances. In a later stage the recoverpoint appliances were also introduced as virtual machines called Recoverpoint/VE.
The technology in itself is so very cool, but unfortunately often complex and hard to do right. Until now. When I first heard if this and figured what it could do, it blew my mind: Recoverpoint can now split the writes from within the VMware hypervisor.
Recoverpoint VM: Splitting writes from inside the hypervisor
First imagine how cool the Recoverpoint solution can be; Recoverpoint splits writes and puts them into a journal. Because we now have all the writes in a journal, we kind of have a snapshot for EVERY point in time; we can roll back the journal and resume a workload on an almost per-write basis. You can also replicate these journals between Recoverpoint appliances to make a vendor-independent very robust replication solution.
Now for the mind-blowing part: WHAT IF… You could split the writes not on the SAN level, but on the VM virtual SCSI bus? This would mean a few things:
- Recoverpoint VM journals and replicates VMs, independent from hardware. FC, iSCSI… Even NFS or local storage;
- Recoverpoint VM will inherently work on a PER VM level. No more LUN level snaps and replication;
- Recoverpoint could integrate with VMware Site Recovery Manager. Recoverpoint VM will make SRM redundant;
- Recoverpoint VM enables replication on a much larger scale than vSphere Replication can;
- As Licensing is on a per-VM basis, Recoverpoint VM can start really really small;
This set of features REALLY impressed me… It lifts so much of the issues we had with replicating the virtual world… and it lifts so much limitations Recoverpoint once had.
Just some bulletpoints to show the true power:
- VM-level granularity Test, failover, failback to any point-in-time per VM;
- Fully virtualized – hardware independent;
- Integrated management into VMware’s vCenter GUI;
- Tier-1 “Enterprise Class” replication solution;
- Builds on 120,000,000 run-hours engine.
How it works on the inside
A picture says a thousands words. In a Disaster Recovery scenario we project two environments with two vCenter servers and the Recoverpoint VM solution:
In the schematic above, you can see that this solution supports ANY KIND OF STORAGE as long as VMware vSphere supports it. Why? Because the Recoverpoint VM Splitter sits inside the hypervisor inside the vSCSI path to the VM.
In this environment we also have virtual Recoverpoint appliances (vRPA’s) that will absorb the writes that were split in the hypervisor and journal and replicate them to the other data center.
Finally, everything is managed straight from the vCenter GUI:
That’s it. That’s all. Where is VMware Site Recovery Manager? It’s not in here… it is hardly needed at this point! With a few simple vCO workflows we could initiate a controlled failover… All the hard parts like talking to the underlying array though the SRA is all gone. All hardware dependency… Gone.
I think this is SO COOL.