A capabilities based communication model for high-performance distributed applications: The Open HPC++ approach

Typical high-performance distributed applications consist of clients accessing computational and information resources implemented by remote sewers. Different clients may have different requirements for accessing a single server resource. A server resource may also want to provide different kinds of accesses for different clients, depending on factors such as the amount of trust between the server and a client. The requirements or attributes of remote access, such as data compression (and encryption) or client authentication can be encapsulated under the concept of remote access capabilities. This paper presents a capabilities based model of communication and describes how it is implemented at application level in a programming environment called Open HPC++. Open HPC++ capabilities are associated with a remote reference and determine the kinds of remote access that it supports. Capabilities can be exchanged between processes, and can also be changed dynamically to helps applications adapt to varying run-time environments. Capabilities also work with the load-balancing features of Open HPC++ to help applications achieve higher performance. The paper presents a set of experiments to demonstrate the usefulness of the capabilities based model. It also describes Open HPC++´s communication protocol adaptivity mechanism, which is the basis of the capabilities based model