Fall 2002. Distributed/Multiprocessor Operating Systems.

Class synopsis
- just a brief outline of topics covered in class

No guarantees of accuracy or completeness.

Class 1: Aug 26

• Introduction to Distributed Operating Systems
• Multiprocessor operating systems
• Problems with multiprocessor machines and differences between conventional and multiprocessor operating systems

Class 2: Aug 28

• Classes of Multiprocessors - UMA, NUMA, NORMA
• Memory contention in Multiprocessor operating system (data, code, OS memory)
• Coherence and caches in multiprocessors

Class 3: Sept 2

• Kernels of multiprocessors, the reentrancy problem
• Locking for preventing race conditions
• Atomicity - interrupt disabling or atomic instructions
• Techniques for atomicity in uniprocessors and multiprocessors

Class 4: Sept 4

• Memory contention in multiprocessors - due to sharing of data
• Using and implementing MUTEX locks
• Reentrancy in kernel and application code for shared memory systems
• Parallel programming using the fork-join construct

Class 5: Sept 9

• Using Barrier Synchronization in Parallel Programs
• Implementing Barrier Synchronization using Semaphores
• Scheduling threads on UMA machines - co-scheduling, affinity scheduling, smart scheduling

Class 6: Sept 11

• Architecture of NUMA systems
• Programming the NUMA machine
• Operating systems for NUMA machines (trivial)
• Parallel Processing using multiprocessors
• The packages - PVM, MPI, Open MP, Linda
• Programming with PVM

Class 7: Sept 16

• Guest Lecture - Prashant Dewan - Domain name Service

Class 8: Sept 18

• Guest Lecture: Austin Godber: Security

Class 9: Sept 23

• PVM, MPI and Open MP, programming
• Shared memory vs/ Distributed memory (for parallel processing)
• Introduction to Linda

Class 10: Sept 25

• Linda - how to program and use for parallel processing
• Linda - execution characteristics (worker pool model)
• Introduction to Distributed Systems and Distributed Operating Systems

Class 11: Sept 30

• The NFS file system
• Mounting and namei - resolving external file names
• NSF client and NSF server operation (including caches)
• Performance issues in NSF

Class 12: Oct 2nd

• The Andrew File system - how to handle scalability
• Distributing Operating Systems (logical/physical and the message possing model)
• Message passing - a conceptual design

Class 13: Oct 7th

• Blocking vs. Non-blocking sends and receives
• Message passing - how to program
• Using ports, exchanging messages, port directionality
• The client-server programming style
• Why asynchronous programming is BAD
• Naming of ports
• Nameservers and well known ports

Class 14: Oct 9th

• Recap of message passing programming strategies (client server is good)
• Using (or not using) timeouts
• Problems with client server programming (bottleneck, failure, recursion)
• How to get around, above problems
• Multithreading the servers

Class 15: Oct 14th

• Implementing message passing - one machine
• Kernel data structures
• Code for send and receive
• Discussion of data structuring, dynamic memory allocation, blocking

Class 16: Oct 16th (homework due: hand in, in class, or via courier from off campus)

• Distributed implementation of ports
• Naming of ports - location independence
• Sending messages to remote ports (netmessage server)
• Receiving messages from remote ports - complicated interaction with netmessage server and proxy ports

Class (noclass): Oct 21st

• MID TERM EXAM

Class 17: Oct 23rd

• Message Passing Operating Systems
• The Microkernel Architecture
• Threads and coroutines
• Using coroutines as threads
• User level threads (intro)

Class 18: Oct 29th 

• User level threads and Kernel level threads
• Multithreaded servers 
• Single threaded stateless servers with cookies

Class 19: Oct 30th  (homework due: hand in, in class, or via courier from off campus)

• Locking in distributed systems
• Advisory and mandatory locking
• Shared and Exclusive locks
• Designing a Lock server

Class 20: Nov 4th

• Upgrading read locks to write locks
• READ http://developer.apple.com/techpubs/macosx/Darwin/General/KernelProgramming/synchronization/Read_Write_Locks.html
• Downgrading
• Merits and demerits of upgrade/downgrade
• Implementing upgrade/downgrade
• Starvation and Deadlocks
• Introduction to RPC

Class 21: Nov 6th

• RPC Continued
• The IDL compiler
• Name service, version numbers and how clients find servers
• Introduction to DSM

-- Nov 11 - holiday

Class 22: Nov 13th

• DSM memory consistency models (click for slides)

Class 23: Nov 18th

• Implementing sequential consistent DSM
• (page based, page fault handling, DSM server, invalidations)
• Page shuttling problem
• Locking and DSM
• Release consistent DSM

Class 24: Nov 20th

• The Muddy Children Problem
• Hierarchies of Knowledge, Common Knowledge
• Consensus in distributed systems (coordinated attack problem)
• Two-phase commit protocol.

Class 25: Nov 25th

• Introduction to Distributed Snaphstots (Chandy, Misra, Haas).
• The paper is here

Class 26: Nov 27th

• Slides are here
• Time synchronization and Event Ordering
• Lamport Clocks
• The paper is here
• (slides are to come)

Class 27: Dec 2nd

• Examples of Distributed Snapshot execution
• Proof of correctness of the algorithm
• Properties of the recorded state.

Class 28: Dec 4th

• Event ordering and logical clocks
• Synchronizing logical clocks
• Distributed Mutual Exclusion
• Lamport's Algorithm for Dist. Mutex

Class 29: Dec 9th, LAST DAY OF CLASS

• Review Session

Dec 17th, FINAL EXAM 10am-11:50am