Operating Systems Spring 2004

1
Operating SystemsSpring 2004

• Summary

2
Important for the exam

• Wednesday 2nd, 845-1245 M building
• You may have with you Operating Systems by
  Stallings or Operating Systems Concepts by
  Silberschatz-et-al
• This is to allow to look up details that may
  escape your memory DO NOT COPY!
• Answers to questions will of course not exist in
  raw form in the book anyway
• You may also have a dictionary (english-swedish-en
  glish)
• No calculators, PDAs, etc (if/where numbers
  matter, you may do rounding)

• Summary study
• accompany your summary study with the notes
  received, exercises (those solved in class and
  others, incl. those in the hard-copy notes) and
  summary questions _at_ the end of each book chapter
• keep a critical eye in your overview study
• why is this so? how does it work?
• How do the puzzle pieces fit together?
• later check os web-page, your os-accounts for
  news

3
Computer System Overview The headlines

• Basic system elements, registers, instruction
  cycle
• Introducing interrupts interrupt cycle
• Using interrupts for IO and multiprogramming
• DMA performance impact
• Memory hierarchy, caches performance based on
  success-rate

4
Operating System OverviewThe headlines

• OS system structure objectives and ways to
  achieve them in each OS sub-area/problem
• Processes/threads
• synchronization,
• scheduling
• resource allocation
• memory
• management allocation
• virtual memory
• file system
• Organization, disk management
• Disk scheduling
• Security issues
• Authentication
• access lists

• Some history
• HWOS co-evolution
• Modern OS issues
• microkernels
• threads
• multiprocessors
• clusters and distributed OS

5
Processes and Threads

• Process a program in execution (unit of resource
  ownership)
• Thread a sequential thread of instructions
  execution- part of a program on execution (unit
  of dispatching)

• Process lifetime (states)
• Process bookkeeping (PCB, tables, queues)
• Incorporating threads in bookkeeping
• User/kernel-level threads
• Examples solaris, java

6
Synchronization of Processes

• Petersons solution for 2-process mutual
  exclusion
• synchronization tools associated with HW support
• distributed mutual exclusion
• Shared Memory
• Lamports Bakery algo
• Message-Passing
• Token-Passing
• RicardAgrawalas solution using timestamps
  (simulating bakery algo)
• timestamps implement the happened-before (a.ka.
  causal) relation to associate events in
  distributed systems
• Some other synchronization problems
  assumed/studied
• bounded buffer (using semaphores, messages)
• concurrent readers-exclusive writers
• dining philosophers,

7
Resource Allocation Deadlocks

• Deadlock Characterization
• How do we deal with deadlocks?
• Prevention requests are placed so that deadlocks
  can never occur (Dining Philosophers as running
  example)
• Avoidance deadlocks are possible, but OS uses
  knowledge of max requests, to grant requests so
  that the system is in safe state (otherwise
  suspends process) Dijkstras Bankers Algo
• Detection (same as Bankers algo?) and recovery
  (process termination, resource preemption)
• Generalized Distributed Resource Allocation with
  Deadlock Prevention

8
Memory Management

• Key issues
• sharing, protection, relocation
• Memory allocation schemes (architecture, hardware
  support, performance issues, fragmentation)
• contiguous
• paging
• segmentation
• combined

9
Virtual Memory, Demand Paging

• Page Replacement
• FIFO
• Optimal
• LRU
• LRU approximations
• Cleaning Policies
• Allocation of frames
• fixed
• variable (replacement local or global) using
• working set model
• page-fault frequency scheme
• Thrashing
• and prevention by variable frame allocation and
  process suspension

10
Process Scheduling

• Long, medium, short term schedulers
• Short-term scheduling criteria/optimization goals
• CPU utilization, Throughput, Response time,
  Fairness, Scheduling overhead
• Short-term scheduling algorithms in uniprocessors
• FCFS
• Round Robin
• Shortest process next (optimal non-preemptive in
  avg. response time), Shortest remaining time next
• determining length of next CPU burst (exponential
  averaging)
• Priority scheduling
• dealing with starvation
• Time-slicing multi-level queues, aging,
• Scheduling Anomalies priorities blocking -gt
  priority inversion convoy effect

11
Process Scheduling (cont)

• Real-time Scheduling for uniprocessors
• Earliest Deadline First dynamic priorities,
  optimal in meeting deadlines
• Rate-Monotonic Scheduling static priorities,
  often used in practice

• Scheduling in multiprocessors
• different degrees of parallelism may imply
  different needs/algos
• Design issues
• global ready-queue or per-processor dedicated
  queues?
• Who executes the scheduler?
• Multiprogramming on processors?
• Approaches
• load sharing (global ready queue)
• gang scheduling (assign threads to particular
  processors)

12
Disk Scheduling - File management

• Why is disk scheduling so important?
• Disk performance parameters seekrotational
  delaytransfer time
• Disk scheduling Algos
• FCFS, Shortest Seek Time First, SCAN (elevator),
  Circular SCAN ((C)Look optimized (C)SCAN)
• File System SW architecture
• logical organization (user interface, file
  organization sequential, indexed, hashed),
• physical organization (block allocation/file
  allocation contiguous, chained, indexed)
• Disk Caching replacement algos
• Virtual and distributed file system concepts
• File sharing access rights, simultaneous access
  (readers-writers per file/block)
• Journaling File systems

13
Distributed OS Issues - Middleware

• Communication
• Client-server model
• Remote Procedure Calls
• Peer2peer concept
• Process Migration
• Clusters
• Architecture, OS design issues
• Relation with SMPs

14
Security pot-pouri

• Types of threats in the system
• Intrusion/Malicious programs detection
• Encryption conventional, public key (quick)
• Authentication (quick)
• Key management
• Guest lecture

15
Instantiation UNIX

• Kernel, system/user programs system calls POSIX
  interface
• multithreading
• Solaris light-weight processes (threads)
  process structures
• Linux only kernel-level threads (clone can be as
  a fork)
• Scheduling
• multilevel-feedback queues (favours I/O bound
  procs) with aging (CPU_bound procs should not
  starve)
• real-time scheduling support preempt-able
  kernel, page locking to reduce page-faults on RT
  processes
• virtual file-system concept

16
Instantiation Windows 2000

• Client/Server computing base for distributed
  computing
• Modified microkernel architecture (but not a pure
  microkernel)
• modules can be moved, removed, upgraded, or
  replaced without rewriting the entire system
• Scheduling policy similar to unix-scheduling
  algo
• Guest lecture

17
Good luck in your exam!

• Have a rewarding vacation time afterwards!