DRaaS-tic action: Trust the cloud to save your data from disaster
In modern computing, disaster recovery can be thought of in the same way as insurance: nobody really wants to pay for it, the options are complicated and seemingly designed to swindle you, but it is irrational (and often illegal) to operate without it.
All the big IT players are getting into disaster recovery as a service (DRaaS), and many of the little ones are too.
The core concept is simple: someone with a publicly accessible cloud stands up some compute, networking and storage and lets you send copies of your data and workloads into their server farm.
If your building burns down or some other disaster hits your company, you can log into the DRaaS system, push a few buttons and all the IT for your entire business is up and running in moments. If only car insurance were that easy.
But like car insurance, DRaaS comes in flavours. There are so many options from so many vendors that the mind boggles.
Prices and capabilities vary wildly. Perhaps most importantly, the amount of effort required to make the thing work properly, and keep it working, can vary quite a bit too.
Simply using software as a service offerings for critical functions and letting the rest burn is not particularly rational either. Public cloud services still need to be backed up.
Vendors go under. Some putz could hack your account and delete everything. A plane could fall out of the sky and land directly on the storage array containing the only copy of your data.
So you cannot avoid disaster recovery planning. You can, of course, set up your own disaster recovery solution. Go forth and build your own data centre, or even just toss a server in a colo.
Both are excellent options, if the circumstances, requirements and budget of the company are right. For everyone else, there's DRaaS.
Turn up the heat
For various reasons (which I have previously discussed), I feel that VMware has the best hybrid cloud offering.
Now, there's certainly some wiggle room in any attempt to classify one service as better than its competitors, and the circumstances of the companies in question absolutely matter.
I doubt very much that a company running 100 per cent Hyper-V is interested in VMware's vCloud Air, DRaaS or not, because it uses an incompatible (for now, anyway) platform; people tend not to care about something that is realistically not an option.
That said, the majority of companies do use VMware for virtualisation. Even where Hyper-V has made a dent in the enterprise, it typically lives alongside VMware's offering. For these reasons I will stick to VMware's DRaaS offering as my example.
The ability to move virtual machines to and from a public cloud without fuss is is the original promise of the hybrid-cloud concept and it is at the core of VMware's vCloud Air. In essence, the company’s DRaaS is just a licensing and pricing wrapper on that service tied to a forked version of its replication software.
Its key features are:
- Self-service disaster recovery protection for virtual machines
- Recovery point objectives from 15 minutes to 24 hours
- Automated failover testing, planned migrations and recovery
- Elastic cloud compute and storage capacity
- Support for offline data seeding
- Private leased-line network options
- Flexible failover testing
Reading between the lines I suspect that the automation portion of the exercise is gelled in the code, tested and ready for prime time. But VMware is not Amazon, Microsoft or Google. There could be some problems if too many Fortune 500s decide to test the failover of 50,000 virtual machines without a little warning,
This gets to the heart of DRaaS and how it differs from vCloud Air's compute-as-a-service (CaaS) offering.
If things start to look a little tight in the data centres, the magic algorithm summons an accountant
While they both use the same infrastructure, vCloud Air's CaaS side offers "hot" compute: resources are reserved for every virtual machine you commit to the service. If things start to look a little tight in the data centres, the magic algorithm summons an accountant to perform the dark hardware-purchasing rites.
DRaaS offers "warm" compute: only the compute required to light up the replication virtual machine is actually committed. You get a block of storage set aside for your use, and that's it.
Failover testing and real emergencies do not happen often enough for VMware to keep idle servers around to handle every single company failing over simultaneously.
Instead, VMware overcommits resources – only the company knows by how much – and uses an in-house algorithm to determine how much idle capacity should be on hand to handle the disaster recovery needs of the collective mass of customers.
vCloud Air is proven tech by now. The DRaaS offering is based on the vSphere Replication software, it too proven and trusted technology.
At time of writing a single VMware DRaaS core subscription can scale up to 500 virtual machines. It should also be noted that VMware commits to completely allocating the compute resources you require within four hours of things going pear-shaped.
How it works
To make VMware's DRaaS work you need vCenter 5.1 or higher and vSphere Replication 5.8. Provision a disaster recovery virtual data centre within vCloud Air, verify that you have network connectivity and you should be good to go.
In this sense, it is really no different from DRaaS offered by any other company. You need to have a subscription on the cloud side, virtualisation software on your side and some sort of widget to do the replication between the two.
VMware already had software to do disaster recovery between two sites owned by a company. During the beta process for its DRaaS VMware forked its software into the standard version, and one specially configured to work with its DRaaS offering.
Having proved itself vSphere Replication 5.8, the latest version, will work with both Site Recovery Manager and vCloud Air DRaaS.
Once you have everything installed, you configure your recovery point objectives service levels. Currently, VMware's DRaaS supports objectives from 15 minutes to 24 hours, assuming you have the bandwidth to push changes as often as you configure them.
The details of how failovers are handled by your cloud service provider matter a lot in deciding which one will meet your needs. While you still find older documentation saying that you are restricted to two failover tests a year, VMware DRaaS customers now get unlimited failover testing as part of the package.
In the event of an emergency you have 30 days to run your failed over instances before VMware makes you pay regular vCloud Air rates for hosting workloads on its cloud. Now I am a cheap git but even I think that's reasonable.
For the smallest of the small, cloudy DRaaS just makes sense. Standing up your own servers is expensive, if for no other reason than that you have to license both sides of that equation.
For the largest customers, just having someone take care of the hassle makes it worthwhile. Quality IT people can be hard to find and you don't want to waste their talents with just keeping the lights on.
It is in the middle of the market that pricing will ultimately tell. This represents the largest number of businesses, but also the greatest diversity of business requirements, risk tolerance and capital availability. ®