After black Friday where American citizens can buy at low price goods in most of the shops comes the Cyber Monday. On that day eShops make incredible discount on their products via the web which attract a lot of cyber customers. This has for consequence a short and huge increase of the load of all servers owned by these companies. This sudden load increase is also called peak load.
Some companies were not prepared to this peak load and thus their infrastructure wasn’t ready to support such load which generates performance degradation and so a frustration from the customers. To avoid this kind of problems, IT architects have to scale the infrastructure for peak loads but the associated cost could be very high as most of the time servers installed to sustain peak load will be under-utilized or idle. Don’t forget that under-utilized or idle servers can cost a lot of money because they still consume power, ground floor, cooler, licenses and human hours for maintenance. As the total cost is very high, IT architects have to, for budget reasons, mitigate the risk and go for an under-sized environment which will not able to manage the peak load or manage it in an acceptable way and/or have special plan to put in place the necessary physical servers few weeks/months before a foreseen peak load.
One of the solution to go smoothly through this peak load situation is the cloud and more specially the elasticity characteristic of cloud. Imagine you have a farm of virtual servers and this farm can automatically adjust its size depending on the load. You will not be only away from peak load problems but you will use only what you need and when you need it. On top of it if you are running on private cloud, this will free up servers for other application and if you are on a public cloud you will simply pay less as you use less cloud resources.
Elasticity is one of the characteristics provided by the IBM SmartCloud Application Services solution. As IBM Workload Deployer and IBM PureApplication are based on the same technology as IBM SmartCloud Application Services, this characteristic is also available on these two solutions designed for private cloud.
[youtube]http://www.youtube.com/watch?v=-cd2xE9hIlo[/youtube]
The elasticity is defined by policy, some pattern-type provides elastic policy out of the box such as the Web Application pattern-type of SmartCloud Application Services. You can define rules based on CPU usage, Request response time and connection pool between your application server and the database. Read this article to have a glimpse of elasticity with IBM SmartCloud Application Services: Is Scalability Your Concern?
Also a Elastic load balancer is available which it is a cluster of load balancers with the capability to add/remove members based on CPU usage.
In conclusion, the elasticity characteristic of cloud can really help in the infrastructure sizing management but of course the application which sits on this infrastructure must be designed to support the elasticity of cloud and we called such application a cloud-centric application.