Single address space OS review

From: ahemavathy (ahemavathy_at_hotmail.com)
Date: Sat Jan 17 2004 - 19:26:25 PST

  • Next message: Reid Wilkes: "Opal paper review"

        The paper describes Opal, a single address space operating system intended to enhance sharing between complex applications that richly interact with each other. It is specifically designed for 64-bit architecture. The authors rightly feel that the various operating systems existing at that time continued to use traditional operating system structures more suited for 32 bit architecture, causing extra work at several system layers. They had built Opal alongside other environments on a microkernal operating system Mach. The Opal is influenced greatly by other capability and object oriented OS systems such as Hydra and Multics.

        The Opal units of storage allocation and protection are segments. A protection domain is an execution context for threads. Threads can access segments or other resources by means of capabilities they possess. A portal is an entry point to a domain. Resource groups take care of resource management. These are Opal abstractions that allow applications to communicate freely with one another.

        I liked the organization of the paper. It has clear and precise summary for each section and a final conclusion. The figures dipicting the three choices of sturctural co-operation between application components, Organization of the opal prototype, tree-indexing using mediators are very good and don't need any more explanations. I liked the fact that the authors have concentrated on just one goal : to make sharing easier and left the other duties of the OS to be managed by the existing OS (Unix & Mach microkernal in this case) I feel that since with the first step they have clearly proved that Opal improves performance on sharing without compromising on security the authors could have gone ahead and built a system from scratch without building on top of an existing system which they say was an impediment. The Boeing example conveyed that the goal of the Opal was achieved.

    PS: Since we haven't finished the class discussion I wanted to submit the review although I am past deadline.


  • Next message: Reid Wilkes: "Opal paper review"

    This archive was generated by hypermail 2.1.6 : Sat Jan 17 2004 - 19:26:59 PST