Cloud If you are wondering why hybrid clouds are the future, look no further than last week's massive crash in Amazon’s Virginia data centre, where its EC2 cloud and related storage clouds are hosted.
Despite the fact that Amazon creates supposedly independent “availability zones” within its data centres to ensure that outages are contained and minimised, something went terribly wrong.
The startups that love the EC2 cloud for its simplicity and usage-based fees, and the enterprises that rely on EC2 for part of their computing capacity, were knocked offline for nearly two days.
The internet was a somewhat quieter place with Quora, Sencha, Reddit and FourSquare all unplugged from the digital ether.
Amazon’s outages were far longer than the four hours of downtime a year that its service level agreement stipulates. Amazon will explain itself to customers and refund some money, but it will surely lose some customers to competing clouds.
More significantly, the outage will make companies realise that they still need a disaster recovery plan for cloudy infrastructure.
The lesson is an old one, but one that IT shops have to keep learning over and over again: if you sell eggs for a living, don't put all of them in one basket when you walk back from the coop.
The Amazon outage will also almost certainly be a boon for the idea of hybrid clouds.
When the cloud hype cycle started a few years back, hybrid clouds allowed companies to use a mix of virtualised server infrastructure both inside their firewalls and on multi-tenant clouds such as Amazon's EC2.
Hybrid clouding also means being able to offload work running on virtualised servers inside your own firewall onto public clouds – what is often called cloud bursting. It’s a kind of temporary outsourcing made possible by the virtualisation of workloads and high-speed networking between the data centre and the cloud provider.
And it is largely theoretical because of the incompatibility of hypervisors, virtual machine formats and cloud management tools.
You can build an internal cloud that mimics Amazon’s EC2, for example, using Canonical's Ubuntu Enterprise Cloud, which supports the Amazon APIs and uses Eucalyptus Systems' open-source cloud fabric. But it’s not as easy as it sounds.
Double or quits
At a recent cloud computing event at Intel's CloudBuilders laboratory in Beaverton, Oregon, the company tried to show how it could move an application being orchestrated by the OpenCloud cloud fabric (created by the Xen project and commercialized by Citrix Systems) from its lab in Folsom, California, to the Oregon lab. The network link between the labs crashed.
No cloud is immune from crashes, whether caused by human error or natural disaster
For hybrid clouds to work, you need multiple network paths between your data centres and your clouds and, very likely, between your various public or semi-private clouds where you shuffle off applications to get some work done.
No cloud is immune from crashes, whether caused by human error or natural disaster. Companies must wise up and double up their capacity – and that means there will be different orders of hybrid computing going on in the IT department.
If you thought cloud computing was going to just let you do your data centre on the cheap, you will be disappointed. But the advent of public clouds supporting EC2 APIs or running VMware's ESX hypervisor gives you the option of spreading your workloads, and provides a level of fault tolerance and disaster recovery that you could never have inside your own data centre.
The IT budget is not going to go down, perhaps, but the service level and performance should go up, and that is enough to keep the bean counters off your back. ®