One of the most important characteristics of a cloud deployments is the ability to scale dynamically when severs seem loaded. Some of the metrics used for scaling include CPU, Memory and Bandwidth utilization. However, in most cases, these metrics are local to a specific system. The dynamic provisioning of additional capacity is also reactive to peak demands. This directly translates to loss of response for a short interval of time when the capacity is being allocated.
Enterprise deployments usually are a collection of heterogeneous system with well studied patterns of stress propagation. These stress patterns usually progress from geographic location to another or from one type of server to another (web server to database, etc). Hence, a provisioning system on a global scale would allow a proactive provisioning system, adding computing capacity to the correct type of servers.
One of the ideas we presented at the RSA sMashup challenge was to demonstrate this dynamic provisioning on a global scale. We picked up RSA envision to collect logs from servers deployed on Virtual Machines hosted by VMWare ESX Server. The trigger that provisions more machines are configured into the reports and alerts at EnVision. The alerts call a batch file that contains VI SDK commands to create and start servers.This file takes care of cloning the machine, bringing it up, etc.
Thus, envision alerts when it notices loads on servers that in turn provisions more servers that are prepared for the load when it progress to them.
