SelfStabilized Operating System Shlomi Dolev BGU Reuven Yagel

Self-Stabilized Operating System Shlomi Dolev, BGU Reuven Yagel, BGU-Rafael

Talk Outline • • Operating Systems & Self Stabilization OS Research What can go wrong? Solutions Top-down vs. Bottom-up Minimal OS Model

Operating Systems & Self Stabilization • Initial Goal: Find the means of making the core(main) components of an OS behave in a Self. Stabilized manner. • Self Stabilization – Usually Distributed Computing Oriented – Fair composition [Dolev] • Operating Systems – Top-down vs. Bottom-up view Virtual Machine vs. Resource Manager [Tanenbaum] – Historical OS main modules: Resource mgmnt: CPU, Memory, Device Drivers & File System.

OS Research • Deprecated by industry & hackers – # of courses in univ. • Base abstraction for higher abstractions (middleware, protocol stacks) • Essential for system reliability. [kernel projects] • Operating System. Net (Ideas will move on to distributed OSs – see sciam article, . Net Framework diagram

What can go wrong? • Everything! • Soft errors. – Approximately 98% of RAM errors are soft errors. • Different OS algorithms.

Solutions • Build on Linux – Open, Modern, Documented, Unix tradition, Cool… – But: “Hackered”, Monolithic – Case study for OS components. • Hardening – Security of servers • SS-Shell – First try, difficult to prove (show code snippet) • SONIX – Reinstalling OS periodically, (Critical Apps? )

Top-down vs. Bottom-up • Top-down – Linux kernel: 2 M source code lines – Windows 2000: ~30 M – Shell • Bottom-Up – We know what we have! – Will take some time to become full operational.

Minimal. OS • IA-32 Architecture – – CPU modes Boot process Memory map, ROM areas Coding & Debugging tips • Watchdog handler – ROM (Liskov’s article) – Program Restart (Von Neuman vs. Harvard models? ) • Demonstration

more • Some model