Techniques for provisioning and decommissioning web servers in an automated manner. In one embodiment, a threshold value relating to an operational metric of a web server can be stored. The metric can be monitored as the web server receives requests. If the metric exceeds the threshold value, additional web servers can be automatically provisioned to handle a portion of the requests. If the metric falls below the threshold value, the additional web servers can be automatically decommissioned. In another embodiment, information can be logged pertaining to web server requests received over time. Based on this information and a threshold value, patterns of high/low server load can be determined and a policy can be generated identifying time intervals during which additional web servers should be provisioned (or existing web servers should be decommissioned). The policy can be used at server runtime to provision or decommission web servers at the identified intervals.