The high availability irnplementations provided by J2EE container manufacturers are based on two basic concepts: replication of services over a cluster and load balancing using static policies. Static policies do not use the internal execution state of each machine in the cluster to decide which computer will be employed by the load balancing system. This feature allows the presence of hotspots - overloaded server. In this paper we present a solution to this problema based on a model of dynamic proactive policies. A dynamic policy uses the cluster's internal execution state to make the decisions about load balancing. The execution state considered by the proactive policy consists ofvariables related to the hardware platform, operating system and J2EE platform where the business application is deployed. Some of the variables considered are: threads loaded, %CPU available, business transaction per second, used memory among others. The load balancing system can take advantage of knowing these variables to make more equilibrated and efficient decisions. The solution is general and could be implemented on any certified J2EE implementation. However the model was implemented on JBoss Application Server 3.0.0.