Thorough research into the nuances of Azure and AWS infrastructure-as-a-service will help you to avoid plenty of pain, according to Elias Khnaser, a research director at Gartner for Technical Professionals.
Speaking today at the analyst firm’s Infrastructure, Operations Management and Data Center Summit in Sydney, Australia, Khnaser said his clients often come to him with problems caused by insufficient attention to detail in the planning phase of a move to the cloud.
He cited the fact that AWS has at least three load balancers, but one can’t handle static IP addresses and can listen to only 20 ports. If you don’t research the service well enough to notice that, he said, you’ll end up in pain. The same applies to assumptions about elasticity: you can’t have more than five AWS virtual private clouds in a single region and if there’s a chance you’ll need more that’s the kind of thing you need to know before you plan.
Khnaser’s talk, titled “Top Amazon Web Services and Microsoft Azure IaaS Mistakes You’ll Want to Avoid”, also pointed out that AWS and Azure now each have three different virtual cloud networking options, each with different qualities and security implications. Skim the fine print at your peril.
And then there’s the matter of how to structure an IaaS account. If you have one per application, you could end up with dozens of accounts and miss out on discounts for high usage volumes. Each cloud also has different ways of billing, different rights for account-holders and the administrators of different resources. Khnaser advised wrapping your head around this stuff before you adopt IaaS, rather than later – not least because your security people need a say in this before you put yourself at risk.
He also counselled developing a backup plan for your cloud, because you cannot assume that promises of reliability equate to recoverability of data or services.
The analyst also illustrated why knowing details matters by mentioning a hack for users of AWS storage: if you upload a lot of files with sequential file names they’ll all land on the same storage device, which means that physical resource will end up under constant load from your users. Randomising file names and/or upload times can spread your stuff around inside a cloud and avoid creating a bottleneck.
Khnaser also recommended remembering that a cloud is elastic. He said he often sees clients under-utilising both cloudy VMs and the ability to scale them down to lesser and cheaper instances, thereby denying themselves one of the main benefits of the cloud! Making sure you subscribe to emails from AWS’ and Azure’s advisory services, and following up on their recommendations, is therefore a no-brainer tip to keep your cloud in fine trim. ®