Hyperactive Messages

Hyperactive Messages are similar to Active Messages; but they are designed with two goals. Like active messages, Hyperactive Messages facilitate efficient communication among computations distributed over a network of workstations. But they also facilitate for the dynamic insertion of code or upgrades to existing code into a computation. If a Hyperactive Message carries data it is processed as a normal Active Message. On the other hand, if a Hyperactive Message carries the code for a service, the code is loaded into the process’s address space and linked to an idle thread. With Hyperactive Messages any instance of a distributed program has the capability to dynamically change its configuration by attaching or detaching services. Such a capability can be used for load balancing and dynamic service upgrading.