Inter-object communication platform in distributed operation system

Abstract

This paper describes an inter-object communication platform in a distributed operation system. The operation system must be highly flexible to manage network elements (NEs) continuously and must be highly expandable to maintain performance independently of the numbers of NEs to be managed. To satisfy these requirements, we adopt a distributed architecture consisting of workstations (WSs) and duplex LANs. Redundant WSs are used so that processing may be continued if there are hardware or maintenance problems. Also, WSs are added, and processes are moved or created on-line to maintain adequate performance corresponding to changes, in the number of NEs managed by the system. Duplex LANs are used to guarantee communication between WSs. With such a structure, the inter-object communication platform must hide the locations of destinations, which may be changed dynamically. In addition, communication must be guaranteed even if problems occur with WSs or the LAN. To satisfy these requirements, we implement inter-object communication, which makes communication units objects. Communication messages are transferred among these objects. An object has a logical ID, which hides the physical structure of the system and object locations. The objects are designed to be identical to object instances of C++, with which our system is developed. This gives the software components high reusability and portability. Also location information, which changes dynamically, is efficiently managed. Thus, an object can communicate with other objects by specifying the object ID rather than the object location. Messages are saved, and transferred to the callee when processes are restarted after problems, and location changes are managed. As a result, problems are hidden from objects. In addition, when message routes are switched to an active LAN, when a particular LAN facility has a problem, objects continue communicating with objects located on other WSs.