When Google’s (News - Alert) Gmail service went down for an hour last Friday, the event sparked an avalanche of news headlines, tweets and curse words. In one way it’s not surprising. Any service like Gmail, which had 366 million visitors in December according to ComScore, would be newsworthy when it went down, but the outage does call for some perspective.
If this is the only hour that Gmail is down within a year’s time, that means the service would be up 99.9886 percent of the time. This is pretty impressive for a service that many people get for free. A glance at a couple of cloud services shows guarantees of 99.9999 percent or six nines of uptime. This level of reliability requires purchase of service level agreements (SLA) that typically guarantee the uptime or credit the account when there is any downtime.
This sounds great, but if you look at the numbers it raises questions about whether or not the downtime is measurable at that level. Uptime of six nines translates to 2.6 seconds per month of downtime. Chances are pretty good that no one would notice a glitch that small in a single month.
The only time that the SLAs are a factor is for the annual hour of downtime that hopefully does not come at a critical moment. For some companies, even that much downtime would be costly. Amazon’s 40 minute outage last August cost the company an estimated $4.8 million.
Most companies however, are not multi-billion dollar online retail giants like Amazon is, so paying extra for an SLA that guarantees six nines of reliability may be wasteful. For companies that process numerous online transactions, like huge retailers or time-sensitive transactions like stock or futures trading firms, an extremely high level of uptime is critical, but for others not so much. It’s all a matter of risk analysis. If the cost of protecting against the disaster far exceeds the cost of the disaster itself, then many SLAs aren’t worth it, especially if they are only a fraction better than what Google provides free of charge.
Edited by Ryan Sartor