How can you avoid cost overruns with infrastructure as a service (IaaS) from Amazon and other cloud services? Cloud usage assessment vendors Cloudability and Cloudyn share tips.
8 Great Cloud Storage Services
(click image for larger view and for slideshow)
Use of infrastructure as service is growing, but do users know what they're doing in the cloud? Two critical analyses would suggest that they don't.
Cloudyn is one of several usage assessment and optimization firms that have been started to serve cloud customers. These new companies often make a basic online service available for free, upping the ante with monthly charges once users discover how much they can learn from the monitoring and diagnostic services.
Cloudyn and another such firm, Cloudability, for example, charge $49 a month for usage analysis and reporting. The two paired up at Cloud Connect 2013, a UBM Tech event in Santa Clara, Calif., to offer snapshots of their customers' cloud practices and how those practices could be improved.
Fifty-eight percent run their cloud applications in an uncontrolled fashion, changing the application frequently but not studying how well it's running on its initially selected server and or figuring out where to run it for optimum cost effectiveness, said Cloudyn CEO Sharon Wagner in the session, "Best Practices in Cloud Optimization."
"Eighty percent of instances are underutilized," he said, suggesting that the IT practice of overprovisioning in the data center has carried over to the cloud. Fifteen percent were overusing their server instances, meaning customers were experiencing delays in getting responses or some traffic wasn't able to access the business's cloud application. A 15% over-utilization rate probably corresponds to a 7% loss of business, said Wagner.
Wagner derived those figures from looking at aggregated data from 450 Cloudyn customers over the last 12 months. Four percent of those customers were spending $1 million or more a year with Amazon Web Services, although most were spending far less.
One of the most striking details to pop out of the study, however, was that 16% of the use of AWS Elastic Block Store, the persistent disk storage associated with a running instance, "are unattached and subject to deletion," Wagner told an audience of about 35 attendees at the session. In other words, the original running instance had most likely been decommissioned but the ESB volume left intact, useless but still being charged for.
Average CPU utilization is 17%. Maximum RAM utilization is 64%, he said. These figures are higher than typical data center studies, which have pegged average CPU utilization in the range of 8% to 12%. Even so, they show a margin where significant savings could be realized.
Although customers reviewing these figures might conclude there is a danger of being charged too much in the cloud, Wagner said the opposite. "Cloud vendors love charging less," he said, posting the slogan on the screen, and adding, "This is not a typo."
Cloud vendors want their services to be an economic alternative to IT operations in the data center and would like a clear economic advantage to emerge in customers' eyes. Unused resources for which customers are automatically charged harm that goal. Amazon Web Services periodically boasts of price cuts, many of them small and only a few of them affecting its core EC2 compute service. Nevertheless, graphs of cloud services can show a steady price decline.
Amazon in particular offers a price advantage to customers who sign up for reserved instances, paying an upfront fee for guaranteed access to a defined amount of service for a year. But 81% of Cloudyn customers' instances are running in the more-expensive, on-demand mode. To use reserved instances, companies must know what their steady state usage is and purchase that as reserved instance capacity, said Wagner. It's actually optimal to purchase a little above the steady state usage so that occasional peaks don't trigger frequent firing up of supplemental on-demand instances to cover the traffic.
Mat Ellis, CEO of Cloudability, analyzed spending by 6,000 companies amounting to $400 million. One of the most interesting details he put on the screen was a chart showing how much of that spending took place in AWS' U.S. East data center complex in Ashburn, Va., as opposed to other regions, such as U.S. West in Washington State or Northern California. U.S. East accounted for the bulk of AWS customer traffic. U.S. East accounts for 66% to 70% of Amazon traffic, according to the chart.
"This hasn't changed over a two-year period, despite three major outages" at the site, said Ellis.
But his main focus was on how his customers spent their money in the cloud, not where they spent it. Ellis said he fired up server instances once himself in U.S. West, as opposed to his usual target of U.S. East, and forgot about them, leading to a charge of $1,600 for resources he never used. No one is immune to setting themselves up for overage charges, he said.
Multicloud Infrastructure & Application ManagementEnterprise cloud adoption has evolved to the point where hybrid public/private cloud designs and use of multiple providers is common. Who among us has mastered provisioning resources in different clouds; allocating the right resources to each application; assigning applications to the "best" cloud provider based on performance or reliability requirements.