Last updated: Jan 1, 2008

This is a list of the key concepts that I cover in the lectures, and that I want to make sure you understand. I base over 90% of all test questions on these topics.

If, after a lecture, you find that you do not have a good understanding of one of these topics, please ask me for clarifications.

If, when studying for the exams, you make sure that you have a good understanding of all of these concepts, you should do well on the tests.

First Exam

Lecture 1 - Course Introduction

• key principles
  • mechanism policy separation
  • modularity/encapsulation
  • intefaces vs implementations
  • appropriate abstraction
• resources (serial, partitionable, sharable)
• resource abstractions (simplifying and generalizing)
• roles and functions of the operating system
  • manage and abstract the hardware
  • relation of os, libs, apps
  • why functionality goes in the OS (vs. in apps/libs)
  • batch vs. timesharing operating systems

Lecture 2 - Introduction to Operating Systems

• viewing services in terms of objects and operations
• service interfaces: ISAs, APIs and ABIs
• threads and processes
  • definitions, differences, uses
  • thread state vs process state

Lecture 3 - Computer and I/O and Architecture

• supervisor vs user mode execution
• traps and interrupts
  • state save and restore
  • vectoring, 1st and 2nd level handlers
• busses, controllers, devices, interconnections
• I/O mechanisms, what they are, how they work
  • polled I/O
  • direct I/O
  • memory mapped I/O
  • I/O interrupts
  • interrupt driven I/O
  • DMA
• smart controllers
• disk geometry, factors influencing disk I/O performance

Lecture 4 - Processes, the user view

• object modules, load modules
  • contents, characteristics, definitions
  • the linkage editing process
• process address spaces
  • text, data, stack segments and their characteristics
  • the process of program loading
  • shared libraries and DLLs, structure and loading
  • process stacks and thread stacks
• registers and process context
  • procedure calls and stacks
  • system calls
• asynchronous exceptions and signals
  • signals as s/w analog to traps
  • signal linkage mechanisms
  • details of call stacking (procedures, traps, signals)

Lecture 5 - Processes, the system view

• process descriptors
  • general contents (state, references to resources)
  • resident and non-resident state
  • (clone vs tabula rasa process creation)
• process/thread context switching
  • general activities and their costs
  • details of the context switching process
  • user vs kernel mode threads
• user and supervisor mode execution
  • system call trap gates, reasons, how they work
  • message based requests, costs and benefits
  • synchronous vs asynhconous, costs and benefits
  • virtual machines and OS emulations
• blocking and unblocking, dispatching and yielding
• swapping, primary and secondary storage

Lecture 6 - Scheduling

• goals responsibilities of schedulers
  • (measuring scheduler performance)
• pre-emptive scheduling, advantages and disadvantages
  • scheduling mechanisms and algorithms
  • starvation, how it happens, how to prevent
  • optimal time slice length
  • priority based scheduling
  • (multi-queue scheduling)
• scheduler behavior
  • (ideal and real throughput and response time)
  • starvation

Lecture 7 - Synchronization

• general classes of race conditions
  • conflicting updates
  • the sleep/wakeup race
  • all or none transactions (to be addressed later)
  • distributed consensus (to be addressed later)
• critical sections
  • what they are, when they happen
  • protecting critical sections with locks
• problems/limitations with disables and spin locks
• atomic instructions
  • what they are, what they do
  • atomic updates in critical sections
  • implementing locks with atomic instructions
• where to lock, advisory vs enforced locking
• synchronous vs asynchronous delilivery and receipt
• the sleep/wakeup-race
  what it is, how it can be solved,
• resource contention and convoys,
  • how they happen, how to avoid,
  • lock granularity, shared locks

Lecture 8 - Synchronization continued

• semaphores and their use
  • how semaphores work
  • semaphores for locking
  • semaphores for asynchronous completion
  • semaphores for producer/consumer synchronization
• correct semaphore implementation
  • solving the sleep/wakeup race
  • (limitations and correct use of interrupt disables)
  • (limitations and correct use of spin locks)
• priority inversion

Lecture 9 - Advanced Synchronization

• monitors and condition variables,
  • why monitors, how they work,
  • why condition variables, how they are used
  • advantages of monitors over semaphores
  • problems with monitors
• the dining philosopher problem
  • how it illustrates deadlock
  • how to solve it with monitors
• inter process communnication
  • shared memory and messages, characteristics and advantages
  • synchronous vs asynchronous operations
  • in-band vs out-of-band communication
  • (flow control)
  • (stream vs msg)
  • (unicast, multicast, pub/sub)

Lecture 10 - Deadlocks

• deadlocks, what they are, how they happen
  • commodity vs general resources
• commodity deadlock avoidance techniques
  • reservations, how and why, graceful failure
  • bankers algorithm, safe over-booking
• four necessary conditions for deadlock,
  • what they are, why each is necessary
  • how to apply each to preventing deadlocks
• live locks, hangs, (hang detection)

Second Exam (first half of final)

Lecture 11 - Memory Management

• responsibilities of memory manager
  allocation, map it in/out, migration
• user address space
• fixed partition, internal fragmentation
  calculating and mitigating fragmentation
• variable partition, external fragmentation
  • variable partition algorithms, and their performance
  • free lists and coalescing
  • (special pools of fixed size buffers)
• garbage collection, what it is, why and how
• the relocation problem
  • what it is
  • how virtual address spaces solve it
  • segment and paged relocation
  • advantages of paging, effects on fragmentation

Lecture 12 - Virtual Memory

• paging
  • paging MMUs and how they work
• demand paging
  • steps of page fault handling
• replacement algorithms and performance
  • spatial reference locality
  • belady's optimal algorithm
  • (stack algorithms)
  • LRU algorithms (clock implementations)
  • working sets, working set algorithms
  • (pre-loading, dirty pages, preemptive write-back)
• paging and segmenation, address space layout
• scatter/gather for DMA transfers
• moving data between kernel and user space

Lecture 13 - Device I/O and Drivers

• device drivers
  • roles and functions, block vs. character devies, special files.
  • classes of drivers and their clients, driver instances
  • DDI, DKI, why they exist, types of functions included
  • dynamic module loading, dynamic vs static configuration
  • driver use sessions, flow of control, execution context
  • layered drivers, why, how organized, how interconnected
• I/O techniques
  • DMA w/completion, chain scheduling, double buffering request scheduling, differences between read and write, solicited vs unsolicited input
  • scatter/gather
  • top-half/bottom-half interrupt handling
  • driver synchronization needs and techniques

Lecture 14 - File Systems (files)

• abstraction of storage, named data w/attributes
• layers of implementation, block I/O, file system integration layer
• levels of indirection in open file reference
• sequential files vs random access
  • mapping between blocks and streams
• keeping track of allocated blocks (general approaches)
  • DOS, Unix, MVS
  • (performance of page finding)
• free lists (general approaches)
  • DOS, Unix, BSD, MVS
  • (defragmentation and performance issues in space allocation)
• indexed sequential files, MVS VSAM
  • structure, motivations
  • record insertion process (and performance)
• volumes and file system structure
  • (DOS, UNIX, MVS volume structure)

Lecture 15 - File Systems (directories and volumes)

• file names, flat, tree structured
• MVS catalogs, DOS directories,
  • absolute and relative path names
• Unix directories
  • links vs copies
  • hard and symbolic links (refrence counting)
• file attributes, operations on containers
  • richess of attributes
• file system performance, block size, cylinder locality
  • LRU block caching and cache performance
  • read ahead and deferred writes
• file system robustness
  • causes of corruption
  • ordered writes
  • audits
  • journaling
• disk partitioning, motivations, FDISK
• file system mounting
• RAID
  • basic types (0, 1, 5) & motivations

Lecture 16 - Protection

• authentication, authorization
  • Unix and DOS authorization mechanisms
• access checking, enforcement, and trusted policy data
  • (ways of guaranteeing enforcement)
  • capabilities and access control lists
• privileged users and trusted software, SETUID (viruses and trojan horses)
• network security, why it is worse
  • man in the middle attacks
  • session keys and SSL
• modern applications of encryption technology
  • public key encryption
  • cryptographic hashes
  • digital signatures
  • public key certificates
  • (certificate authorities and delegated trust)

Lecture 17 - Networking

• Interaction Paradigms
  • work-group computing
  • client-server computing
  • peer-to-peer networking
• implementations
  • layered protocol implementations
  • performance issues and approaches
• challenges of distributed computing
  • Deutsch's seven falacies
  • leases and revocation
  • atomic transactions
  • three-phase commits
  • distributed election protocols
  • cluster membership
  • heart-beating

Lecture 18 - Distributed Filesystems

• classes of solutions
  • (copying, remote disks, explicit, implicit)
  • partitioning the work, client/server architecture
• goals and issues in remote file systems
  • performance, robustness, synchronization, propagation
  • authentication and authorization, [peer-peer vs client session protocols]
  • caching, transactions, stateless servers, idempotent ops,
  • centralized synchronization, server fail-over
  • partition/rejoin, disconnected operation
• examples: NFS, Andrew, HTTP, CIFS
  • design centers, characteristics, advantages, disadvantages
  • layers of implementation, client/server architecture

Lecture 19 - Distributed Computing

• goals for distributed systems
• classes of distributed computing systems
  • SMP, NUMA
  • SSI clustering
  • loosely coupled systems, horizontal scalability
  • motivations, architectures, challenges for each
• distributed computing paradigms: network aware, RPC, object
  • basic RPC model, procedure to RPC mapping
  • local vs. remote objects, issues and solutions
  • Obeject Referenc Brokerss
• evolution from ABIs to protocols
• effect of distributed computing on system architecture