Proactive resource allocation based on the usage pattern (compliance factor)
Original Publication Date: 2005-Feb-17
Included in the Prior Art Database: 2005-Feb-17
In an On Demand world it is critical not to waste any resource and to allocate resources to a system only if the user of the resource complies to predefined rules. On Demand users generally sign a ‘usage’ agreement specifying that they will use the On Demand resources only for specific activities (i.e. porting, testing, certification). The policy based controls in this area are designed to optimize On Demand system resource utilization by giving 'compliant users' higher prioritization for system resources. Users who do not follow the 'rules of engagement' are naturally encouraged to become 'compliant users' in order to gain acces to greater allocations of system resources. For example, a user who ‘fully’ uses the resources (not necessarily in term of CPU usage but also in total duration over the engagement period, ...), or makes the best use of the resources, during the reservation is rewarded by having a higher priority access to resources when needed. The invention defines a software framework that controls the allocation of resources based on the user’s behavior and his/her compliance to the ‘usage agreement’ they accepted before getting access to the resources. The first step of the invention is to translate the ‘usage agreement’ into policies with attributes that can be evaluated and monitored. Second step is to generate a ‘compliance factor’, based on the result of the monitoring, that represents how far or how close is the user’s activity compared to his entitlement. This value is then used for two different purposes: the first one is to adjust dynamically the use of the resources and the second is as a ‘memory’ over a longer period of time (cross engagements) to rank the user and associate a ‘class of service’.