Cloud customers keep learning the rudiments of services the hard way. Case in point: Engineers at a small ISP realized recently that the virtual servers they signed up for weren't backed up by the same cloud service that provided them.
"I am appalled that [our cloud provider] would offer something called 'cloud' without having any failover at all," complained one engineer from the ISP on an industry message board. He said his company's servers were out for hours as a result of not knowing the fine print on the contract.
But he was quickly set straight by other engineers on the email forum, who piled on with the "news" that expecting automatic redundancy from a infrastructure-as-a-service vendor is unrealistic. It's up to the customer to anticipate failures and prepare for them either within its own datacenter or by contracting more servers -- and possibly more support and configuration tools -- from the cloud provider.
"Just because it's 'in the cloud,' [a cloud] doesn't gain higher reliability unless you're specifically taking steps to ensure it," wrote one forum participant. "Most people solve this by taking things that are already distributable (like DNS) and setting up multiple DNS servers in different places -- that's where all this 'cloud stuff' really shines."
Someone else pointed out that adding redundancy or failover to a cloud service gets expensive. He stated that his company uses Amazon services and is happy about it, but had to arrange its own failover operation, albeit using Amazon's tools. "It's up to you to make it happen," the engineer wrote. Further, getting good support options (i.e., getting someone on the phone) costs extra and is "tied to how much of their resources you are using."
Creating redundancy on one's own network can also be costly, especially if storage area networking is used. But one ISP engineer noted that his company simply "swaps [virtual servers] with another ISP outside our geographic area" in order to achieve redundancy. And someone else suggested that since a DNS server was part of the original poster's cloud setup, simply tweaking the DNS records to allow the host to work around a failed host is also an option.
In any event, it's clear from this exchange that cloud services aren't always the complete package they appear to be. IT must do the diligence to establish exactly what is being offered by a cloud provider versus what's needed. Sometimes extra work and/or costs will have to fill the gap.
As one engineer posted: " 'Cloud' outside of references to mists and objects in the sky is a completely meaningless term for operators. In fact, it has made it harder to differentiate between services (which I'm sure is the point)."
— Mary Jander , Managing Editor, Internet Evolution