DR to VMware Cloud on AWS with Site Recovery Manager

VMware has really embraced cloud in a big way. Not that it had a choice, but its decision to abandon trying to build its own public cloud with vCloud Air has been a good one. Freed from attempting to be more AWS than AWS (never do this), VMware can concentrate on what it does well, which is providing a familiar management environment for enterprises to manage complexity at scale.

Moving things around has always been a massive, yet boring, problem in enterprises. Anyone who’s done a data migration project will know the pain of which I speak. VMs are supposed to make that easier, freeing one from the constraints of hardware to a large degree, yet the general mayhem caused by negotiating an outage window for an application makes the technical ease somewhat moot.

VMware Site Recovery makes VMware’s standard Site Recovery Manager (SRM) product work with VMware on AWS by using vSphere replication to move VMs from one cluster to another; the second cluster just happens to run in AWS now. This means it works just like it does now if you have two vSphere clusters.

A big challenge for me is the minimum footprint required as a target for DR to receive the data stream. Because the process uses SRM, the target needs to be running vSphere so it can receive the data stream and write it to storage. That has a minimum footprint of running servers that you have to pay for, as well as enough storage to store all the replicated data. There are some nuances to the specific configuration needs here as well that you’d need to explore, such as how your storage is attached in replication mode and how it gets re-attached to new VMs in DR mode.

The target environment is still smaller than a full physical DR replica of production (while in replication mode), but it’s larger than other methods of doing DR to cloud such as those that send VM backups to S3 and can spin then up on demand. Still, the operational ease might make up for the likely difference in cost, and one needs to view this kind of thing holistically for an organisation.

Deciding whether or not to use VMware native SRM for DR or some other method is the kind of thing enterprise architects look at for whole-of-site protection for critical applications. There are economies of scale to consider, and those economies can be “I trust that this will actually work first time when I need it because everything is on fire.” Brittle hand-crafted scripts are not what you want to use here.

This is really about VMware continuing to enable enterprises to keep their existing processes, for the most part, and just add a datacentre location that happens to be ‘cloud’. It’s a step towards cloud, because you do use the on-demand provisioning of VMs feature of cloud which saves you renting a bunch of empty servers that just sit there as insurance, but you haven’t changed very much about how you operate your IT.

For enterprises with a big investment in VMware, this is a feature, not a bug.

Yes, you get a better overall result if you build your infrastructure for cloud-style operational maintenance, but that’s really quite hard and a lot of organisations have applications that were designed and built when everything ran on a single physical server under someone’s desk. Building applications to operate in a cloud-native way is such a monumental shift in thinking that few organisations are able to do it well.

Don’t overlook the benefit of using this mechanism to vMotion VMs out of a physical datacentre into AWS, even if just as a stepping stone to moving them back into a physical datacentre somewhere. Using rentable swing space like this could be extremely handy as a data migration tactic. Migrating data around the place is tedious and boring but it happens all the time and making boring maintenance tasks easier to do is nothing to be ashamed of.

VMware should be congratulated for realising that it needs to concentrate on being the VMware its customers want, and not trying to be a poor imitation of AWS. VMware customers want different things to AWS customers, even though they are sometimes the same people. Sometimes I like to walk, and sometimes I catch a tram or train. I do not want to put a treadmill on a train so I can walk while I train.

Figuring out which company is selling a treadmill on a train I leave as an exercise for the reader.