Management Strategies For The Cloud Revolution

In this first of a two-part book excerpt, author Charles Babcock, editor at large with InformationWeek, brings shape to the 'amorphous cloud'

Charles Babcock, Editor at Large, Cloud

May 8, 2010

4 Min Read

In one 24-hour period, the site had 41,466 visitors, each staying on the site for an average of six minutes and viewing an average of four pages, for a total of 168,873 page views. Going into the project, Parker had conceived of relying on a single dedicated server to host the site. Rackspace Cloud's general manager, Emil Sayegh, was able to marshal dozens of servers -- up to 100 were serving the site at certain times -- during periods of peak demand, because the combined traffic to all sites in the cloud is automatically monitored, and managers keep a constant surplus of capacity on standby at all times. As Extreme Makeover ate into that surplus, Rackspace fired up more servers to stay ahead of demand. The elastic cloud expanded to meet the need as it materialized.To Parker's surprise, traffic was steady as the show aired Sunday evening and dropped off soon afterward. The spikes had been prior to the show because participants in the project, their families and friends, and interested onlookers were anticipating what the show would reveal and wanted to be the first to know.

Parker says that the experience has convinced Cybernautic to drop its current form of computing, renting servers from a service provider, and to move his clients' 200 Web sites into the multitenant Rackspace cloud, where customers share servers but always have enough capacity to go around.

"I don't need to worry anymore about whether I need to add another server. The cloud automatically scales to match what I need," Parker said.Much of the excitement about the cloud reflects this understanding of its elastic nature, its ability to scale up and down for nearly any kind of business. Building elasticity into the corporate data center used to be handled by buying and installing servers until you had more capacity than you actually needed -- an expensive proposition. With the cloud, it's suddenly possible for a small company, like Cybernautic, to do everything that a big company would do. Best of all, you pay only for the resources that you use, as opposed to buying and installing resources that you might use but that will sit idle most of the time.Amazon charges 8.5 cents an hour to run a Linux server and 12 cents an hour for a Windows server; Rackspace charges 1.5 cents an hour per Linux server. Additional charges are incurred for load balancing, auto scaling, and so on. In general, cloud prices are deemed to be as low as or lower than the charges arising from the most economical corporate data center operation. The economies of scale built into the cloud give it an ability to adjust on the fly for end users who need it -- both large and small businesses -- and an ability to maintain low charges.If more resources are needed, the chances that the cloud can summon them are high. It would be expensive for cloud providers to keep massive amounts of surplus facilities sitting around unused day after day waiting for a rare spike. But the cloud doesn't have to do that because it is a multitenant facility, with many customers using the same servers, and in some cases the same software. The cloud managers make an educated guess at how much surplus capacity is safe to maintain; their advanced load-balancing systems can anticipate need, adding more servers for more direct power, while at the same time moving workloads around to underutilized servers.The cloud's operations managers base their estimate of what constitutes a safe surplus on the analysis of server logs and historical patterns from monitoring the servers' total workload. Managers also hope that not every cloud customer will create a major spike at the same time, an admittedly rare possibility. This appearance of expandable capacity for any single end user is to some extent an illusion. Somewhere, as with all material things, there is a limit to how many major spikes could be met at one time in any given cloud data center. But with thousands or tens of thousands of customers, what is the likelihood that the cloud provider will experience a surge in need from a majority of its customers at the same time? In the normal course of operations, multiple customer spikes are infrequent and, fortunately, spikes rarely travel in herds if the customers are varied in their business makeup. A cloud's monitoring system can show warning flags, send alerts, or sound alarms; somewhere behind the automated system is a management console with a human watching. However, the cloud's own monitoring system is capable of anticipating need and quickly firing up additional "virtual machines" as needed. By doing its computing in the cloud, a business like Cybernautic can grow its Web site business quickly, without hiring a burgeoning IT staff or undergoing the expense of constructing an overbuilt data center with a large margin of surplus capacity. At the same time, a large enterprise could siphon off demanding but noncritical jobs in its data center by sending them to the cloud and reducing the capacity that it needs to keep on hand.To come in Part II: Why the cloud is different from what's gone before, and the high cost of market entry for cloud providers.

About the Author(s)

Charles Babcock

Editor at Large, Cloud

Charles Babcock is an editor-at-large for InformationWeek and author of Management Strategies for the Cloud Revolution, a McGraw-Hill book. He is the former editor-in-chief of Digital News, former software editor of Computerworld and former technology editor of Interactive Week. He is a graduate of Syracuse University where he obtained a bachelor's degree in journalism. He joined the publication in 2003.

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like

More Insights