When planning a web-based service rollout, one of the important things to remember is to plan the capacity of machines that will handle the demand. It is not uncommon for capacity planners to use peak load criteria to size the machines. The implications of using peak load data for capacity sizing are enormous. For instance, a retail organization has to size computing power for the load they expect during the holiday shopping season in late November and through December. The graph below shows the traffic history on Walmart.com (source: Alexa). Say they size the machines and computing capability for the holiday season – what happens during the off-season? Typically, the machines and computing infrastructure lie idle. It is therefore not uncommon to see IT data centers being utilized at 10% capacity during off-peak seasons, incurring unnecessary costs.
This story is not an uncommon one. In 2007, both Sears and Macy’s sites collapsed under the traffic of Black Friday (Pingdom, 2007), and 2008 welcomes the dawn of the “Twitter Effect,” as Mashable’s tweet by Pete Cashmore shuts down a blog site (Pingdom, 2009). Other sites have struggled to stay afloat amidst heavy popularity, and some have paid dearly for it (Pingdom, 2009). , an alleged Google competitor, launched and crashed under heavy traffic, scarring its reputation in the eyes of users (Pingdom, 2009). Europeana, an effort to post artwork from major European museums online, received over 10 million hits per hour, forcing the site to revert to a limited-capacity beta (Pingdom, 2009). Microsoft’s turned out to be more popular than expected, and Microsoft was left scrambling to collect more hardware to manage the traffic, noting, “traffic has far exceeded even our most optimistic expectations,” (Pingdom, 2009). Planning for peak demand invariably results in underutilization of infrastructure. So what should website capacity planners do when demand is unpredictable?
This is where cloud computing infrastructure from someone like Mosso|The Rackspace Cloud comes to the rescue. A blog post on Gigaom.com sums up the 10 laws of Cloudonomics very well (Weinman, 2008) . Elastic capacity, enabled by rapid provisioning, reduces the need to forecast website traffic for capacity planning. Aggregating demand for multiple websites smoothens the total demand as seen on this chart (source:Alexa). The demand from multiple websites seldom goes up and down at the same time due to cyclical variability, thus canceling out some of the fluctuations in aggregate.
Cloud customers benefit from economies of scale, such as volume purchasing, network bandwidth, operations, and administration when a cloud provider like Mosso | The Rackspace Cloud handles these operations. Average unit costs of computing are reduced because fixed costs are spread over more units of capacity and utilized by more tenants. One big reason to use Infrastructure-As-a-Service from a company like Rackspace is that users DON’T utilize 100% their computing capacity 24/7/365, as shown in the chart above. The Rackspace Cloud allows buyers to add compute capacity in mere minutes, as it is needed, while paying only for capacity used. This has a huge impact on the economics of service. Add to this, the support that a company like Rackspace provides to alleviate the needs for IT infrastructure management overhead costs. Even Uncle Sam and the US Federal Government have realized the benefits of these economies of scale, when they declared last month in the that they would be able to cut down $6.6B in costs due to the use of Cloud Computing and SaaS (News Report, 2009).
In general there are several economic advantages to using cloud computing infrastructure:
· Direct costs savings on infrastructure acquisition – Organizations do not have to purchase full computing power; they only pay for what they use.
· Increased agility to introduce new products and services and add new revenue streams – They do not need to worry about setting infrastructure up, as it is already set up and ready to go.
· Increased ability for redundancy, to prevent business losses due to poor performance or physical server crashes.
· Elimination of IT operations and maintenance costs for computing equipment, software, energy, real estate and staff.
Please let me know of any thoughts you have on when and where the cloud makes sense from an ROI standpoint for your company. You can always reach me at firstname.lastname@example.org.
Emil Sayegh; General Manager
Mosso|The Rackspace Cloud
News Report. (2009, February 10). “IT Bailout Report” Identifies Potential Federal Savings. Retrieved February 17, 2009, from Government Techonology:
Pingdom. (2009, February 3). Dawn of the Twitter Effect. Retrieved February 19, 2009, from Royal Pingdom: http://royal.pingdom.com/?s=twitter+effect
Pingdom. (2009, January 5). Online Launch Troubles and How to Avoid Them. Retrieved February 19, 2009, from Royal Pingdom: http://royal.pingdom.com/2009/01/05/online-launch-troubles-and-how-to-avoid-them/
Pingdom. (2007, December 27). The Major Incidents on the Internet in 2007. Retrieved February 19, 2009, from Royal Pingdom: http://royal.pingdom.com/2007/12/27/the-major-incidents-on-the-internet-in-2007/
Weinman, J. (2008, September 7). The 10 Laws of Cloudonomics. Retrieved February 17, 2009, from GigaOM: http://gigaom.com/2008/09/07/the-10-laws-of-cloudonomics/