Fine-grained mobility in the Emerald system

Reviewer: Ste´phane Leonidas Tsacas

.abstract Emerald is an object-based language and system designed for the construction of distributed programs. — From the Authors' Abstract This paper is divided into five sections. In Section 1, the introduction, the authors recall the advantages of process migration and fine grain mobility. Section 2 is an overview of Emerald. Using examples coded in Emerald, the authors show how an Emerald object is described and how abstract types are defined [1]. They explain the set of mobility primitives and a new parameter passing scheme ( call-by-move), and give a description of a process attached to an object. Section 3, titled “Implementing Mobility in Emerald,” is devoted to the description of the data structures, algorithms, and protocols involved in the kernel. Some significant differences between Emerald and other systems are given below. There are three object implementations (global, local, and direct) with separate addressing options. Emerald uses an algorithm for object searching based on Fowler's forwarding address [2], modified for efficiency. It uses direct pointers (as opposed to indirect references as in Smalltalk). Process moving is possible without code copying. Emerald uses machine registers for efficiency (and has an algorithm for moving them). There are two garbage collectors, local and distributed, which can work in parallel with Emerald processes. Section 4 covers performance. At the time this paper was written, the system worked on a network of MicroVAXes connected by a 10 megabit/second Ethernet. A newer version runs on Sun workstations [3]. Tables summarize the execution times for the mobility primitives and the network traffic in the Emerald Mail System. Section 5 is a summary of the paper. This paper is well written. It contains many references covering a large amount of work in the area. The authors discuss the implementation choices and include comparisons with other schemes. There are examples and figures, and the text is clear. Nevertheless, some knowledge in the fields of distributed computing and object-oriented languages and their implementation is desirable. This paper should be read by anyone involved in the construction of distributed systems. Other papers about Emerald have already been published (see [4,5]).