As a part of my job, I work with large enterprise customers trying to help them realize the potential of Cloud Computing. As much I believe in the benefits of Cloud, I also have a realistic discussion around the anti-patterns for Cloud and why certain applications may not be the great candidates to be moved to the Cloud. It makes it easy to have certain patterns in mind when discussing what kind of a line of business application can move to the Cloud for realizing the best return on investment.
In this article, I want to discuss the possible scenarios for getting the maximum value out of Cloud.
Periodic Processing – Every enterprise has at least one application that goes through the grind during a certain time of the month, quarter or year. A classic example of this pattern is employee performance review application. This app is not so much in use for most part of the year and becomes the most accessed and used application during the mid- year review or annual review. In the finance industry, Credit Card companies run their risk scoring during the month end. July is the busiest month for tax filing applications run by Chartered Accountants in India. The common pattern across all these applications is that they are inactive for most of the times but become the most resource demanding applications during a specific period. From the IT investment perspective, it doesn’t make sense to allocate resources for these applications that are inactive for most of the time. Instead of provisioning and de-provisioning the infrastructure, it is optimal to move such applications to the Cloud. If you have such an application, think how you can move it to the Cloud.
Start Small, Grow Fast – This is the proven mantra for many startups. Majority of the startups would be intimidated by the overwhelming cost of the infrastructure. Even within the enterprise, there is a certain class of applications that start small and after testing the waters, they either get killed or become mainstream. Classic examples of such applications are contests and new business processes. What starts as a pilot within a department may turn into a best practice that gets adopted by the entire organization. Since most of these applications are designed from the ground up, they can be provisioned on the Cloud from day one and based on the demand, it can be scaled further. If the pilot is not well received by the employees, the application can be de-provisioned with minimal cost.
Unpredictable Burst – This pattern is more commonly seen in consumer facing web applications. If a website’s link appears on Slashdot or New York Times, there is an instant deluge of traffic and unable meet this demand the website might just become unavailable. If the enterprise launches a new digital marketing campaign which becomes extremely popular and gets viral, there will be a sudden spike in the traffic. These digital marketing campaigns are the ideal candidates for the Cloud. The biggest advantage of moving to the Cloud is the auto-scale capability. By monitoring the patterns in the traffic and the utilization of the infrastructure, additional server resources can be dynamically added to the application. Once the traffic stabilizes and the environment is not so demanding, the extra servers that are added can be safely disposed. This ability makes Cloud Computing extremely lucrative.
Predictable Burst- Ecommerce sites experience the predictable burst all the times. When Apple opens up pre-ordering for their next gen phone, they get ready to handle the peak loads. Because of the predictability, it is fairly common for the business to have a ball-park budget figure of the infrastructure cost. In public sector, there are many events that experience predictable bursts. Electoral results, examination results and publishing of latest census reports will result in this pattern. Sports sites know this better during a popular event like the world cup. Within the enterprise, these patterns are visible during a virtual company meeting or introduction of a mandatory online training for all the employees. The difference between this pattern and the first pattern, Periodic Processing is that there may not be a periodic recurrence of this pattern. The utilization of the resources is influenced by external parameters but the IT decision makers are usually aware of this burst ahead of the actual deployment.
What is your experience of moving an application to the Cloud? If you think there are more patterns, please do share with me by dropping a comment.
[This article was originally published in the Cloud Computing Journal]
