A recent survey suggests that a lot of confusion exists about cloud disaster recovery. Another finds that Disaster Recovery as a Service (DRaaS) uptake will be propelled by “Dynamic industry verticals.” To help get the facts right we’ve enlisted the services of Richard Davies to clear up any confusion that exists and examine the drivers and perceived pitfalls behind its implementation.
Why are businesses interested in DRaaS?
I think that the basic concept of cloud disaster recovery sounds very good because cloud is meant to be on demand. Instead of having a DR site that we pay for 27/7, let’s have one we don’t pay for on this basis. So the headline premise is appealing.
Sounds good, but surely it’s not as straightforward as that?
There are two areas of confusion: one area is that the reality of recovering into a public cloud environment can be more complicated in terms of software set up, etc. Your data centre may have specific IP addresses, and be accessed via various virtual private networks (VPNs), and while you can use them in a cloud environment you have to spend time configuring them.
The second thing is that the traditional cloud model of on-demand doesn’t fit well with DRaaS. For example if you buy an 8Gb virtual machine (VM) and you will pay for that capacity even if there is no load running on it. This means you end up having ‘cold spares’ which you don’t pay for because they are turned off, but this doesn’t fit with the model of the on-demand cloud.
If you want better DR with a replicated environment, then this is more difficult and cost more in a traditional cloud environment. So you need to think more about how you set up your software for a cloud environment.
How does cloud-based disaster recovery compare in terms of advantages and disadvantages to traditional methods of disaster recovery?
The primary advantage is cost-savings, because you don’t have to have a physical site that you pay for 24/7. The primary complication is getting your software to work in the cloud. Additionally with a container-based service such as ours the costs are based on the utilisation of the virtualised server and not its capacity. For example with a Linux container it’s possible to create cloud services that only charge for what’s used. So if you use all of the 8Gb VM you pay for it, but if they just use 2Gb of the 8Gb server then you just pay for the 2Gb used. For more on the savings from containerisation see Richard’s article (Why Linux Containers are the new building blocks for cloud in 2015)
The appeal of cloud-based DR is strong; the reality is that getting your software to run on it will take time, and with virtualised cloud the promise of reduced costs is not guaranteed. The Linux approach means that you are able to use hot spares, but because they are spare the price of running them is relatively cheap. A server in a conventional cloud environment that is turned off, costs you nothing. If it’s on it usually costs you 100%.
Is security an issue with cloud disaster recovery?
If people are moving to cloud-based DR they have to put a bit of effort into thinking about how they are going to set up their software and working out best practices. You have to put in a bit of planning, but it doesn’t decrease security. For example, you might have servers in your own data centre which can be accessed by people there, but if you put it into the cloud you are going to have to do some firewalling.
How do you calculate recovery times?
A business needs to look at risk-cost equations to decide this. You need to ask yourself what will be the impact on my business if the system isn’t recovered in a few minutes, hours or days? If you have systems that are essential and generating revenue, supporting your business and you need them minute-by-minute, then you need look at hot-spare DR. If there are systems you can live without, then you would just need some cold-spare DR.
You need to look at the applications you run and analyse what is critical. Is the e-commerce website critical, for example? Most businesses will have a variety of systems, and customers may place orders through an e-commerce website. Other systems look after industrial processes. For some applications the risk-cost ratio is not worth it.
You have a number of different choices in cloud DR and each one has different costs. A cold spare will need software updates once turned on, and hence will take more time to be up and running, but it will cost less. Hot spare costs more, but it will be more readily available.
Richard Davies, is CEO and founder of cloud Infrastructure as a Service provider, ElasticHosts. Previously to Elastichosts Richard worked at management consultants McKinsey and Forbidden Technologies.