IBM Israel Research Seminars

Interest in robust system design has increased in recent years. One reason is the increase of vulnerably to soft-errors. We suggest dealing with such threats by designing a system to be self-stabilizing. The self-stabilization property captures the desire to recover automatically from any (unexpected) state. A self-stabilizing system converges to a legal execution once faults stop occurring. The concept of self-stabilization addresses the automatic failure detection and automatic recovery facet of, e.g., the autonomic computing grand challenge.

We present new concepts for a stack of tools that provide a software platform for executing self-stabilizing programs: a self-stabilizing processor, a self-stabilizing operating-system ,and a self-stabilization preserving compiler. An elegant composition technique of self-stabilizing algorithms is used to show that once the underling microprocessor stabilizes the self-stabilizing operating system (which can be started in any arbitrary state) stabilizes, then the self-stabilizing applications that implement the algorithms stabilize.

One approach in designing self-stabilizing operating system is to consider an existing operating system (e.g., Microsoft Windows, Linux) as a black-box and add components to monitor its activity and take actions accordingly, such that automatic recovery is achieved. The other extreme approach is to write a self-stabilizing operating system from scratch. An operating system kernel usually contains the basic mechanisms for managing the hardware resources. The classical Von-Neumann machine includes a processor, a memory device and external i/o devices. Thus the tailored operating system is built (like many other systems) as a kernel concerned with managing these three main resources. The usual efficiency concerns which operating systems must address, are augmented with stabilization requirements. Our proofs and prototypes show that it is possible to design a self-stabilizing operating system kernel.

This is a joint work with my advisor Shlomi Dolev.

About the Speaker
Reuven Yagel is a PhD student at the Department of Computer Science of Ben-Gurion University. He is also at Rafael since 1998, where he worked programming and leading several C2 (command & control) and EW (electronic warfare) projects.