As Kent Erickson said in his recent post, What You Going Do when the Cloud Gets Slow on You, moving to the cloud doesn’t give you freedom to tune out the physical infrastructure behind the magic. If you’ve kept up with this blog, you know that pointing fingers at your provider (whether it’s in-house or something like Amazon EC2) when an outage occurs is both unproductive and makes you look like a joker.
So address the physical aspect of your cloud, whether you’re building something internally, outsourcing to a cloud services provider or using a hosted application. Info-Tech lead analyst John Sloan told me that while you don’t necessarily have to tour the physical data center, as you might when looking at colocation solutions, you want to do some due diligence on the kinds of facilities being used and where the hardware actually sits.
“You want to know if this stuff is located in a high-rise or a bunker or whether it’s in an earthquake or hurricane zone,” Sloan said. “Also you want to make sure that things like replication between geographic sites and redundancies are in place.”
Once you’re confident that the physical data center supporting your cloud isn’t sandwiched in an unventilated storeroom underneath the La Brea Tar Pits (to use an extreme and unlikely scenario – but you never know!), you then want to nail down those SLAs, so that you’re covered in case of an outage.
“Infrastructure providers typically have a range. Some of them have meticulously spelled-out SLAs, [while] other ones are [x number] of 9’s or ‘best effort.’ They’ll do their best to not have an outage, but that’s not the same thing as guaranteeing some sort of compensation if they do,” said Sloan.
Sloan’s advice? Take a case-by-case approach. “Ask the sort of availability they guarantee and their SLAs. And, if you’re using an outside provider, have a look at where those facilities are located, whether there are multiple sites and whether that [redundancy] is part of the deal.”
Image courtesy of Joost J. Bakker