MET TC670 B1 Computer Science Concepts in Telecommunication Systems

1
MET TC670 B1Computer Science Concepts in
Telecommunication Systems

• Fall 2003

2
Lecture 6, October 28,

• Midterm Recap
• Multimedia Operating Systems Concepts

3
Midterm Statistics

• Mean 63, Medium 63

4
Problem 1 Processes
0xFFFFFFFF
stack (dynamic allocated mem)
3 points
i
SP
address space
heap (dynamic allocated mem)
static data (data segment)
4 points
hello
code (text segment)
ilt100
3 points
PC
0x00000000

• main()
• int i
• for(i 0 i lt 100 i) printf(hello)

5
Problem 2 CPU Scheduling

• First part easy (5 points)
• Second part 2 possibilities
• New job is put in the ready queue first
• New job immediately runs

6
Problem 3 Synchronization

• Solution
• Spin-locks is a low-level locking mechanism to
  synchronize processes. In spin-locks, the caller
  busy-waits or spins for a lock to be released
  (2 points). Spin-locks is not appropriate for
  uniprocessor systems, mainly because it is
  implemented in a less efficient way. One process
  will run at any time. However in a multiprocessor
  system, efficiency is less critical since often
  there are idle processors. On the other hand, the
  simplicity of spin-locks is very desirable (3
  points).
•  
• Monitor is software module that encapsulates
  thread data structure and procedures in an
  exclusive fashion. Monitor protects the data from
  unstructured access (2 points). With Hoare
  monitor, the waiter runs immediately and its
  condition is guarantee. Hence, it is efficient
  since the waiter does not have to re-check the
  condition. However, the signaler must do extra
  work to restore monitor invariant before
  signaling. With Mesa monitor, the signaler
  continues. Therefore it is possible that when the
  waiter runs again, the condition is changed again
  (so it needs to re-check the condition). Mesa
  monitor is easy for the signaler since it does
  not have to restore invariant. The signaler may
  simply run until out of the monitor session (3
  points).

7
Problem 4 Deadlocks

• Solution
• Four conditions of deadlocks are (1) Mutual
  exclusion, (2) hold-and-wait, (3) no preemption,
  and (4) circular wait (5 points). In this
  example, condition (1) holds as two vehicles
  cannot take the same slot. Condition (2) holds as
  every vehicle holds a slot and is waiting for the
  slot ahead to be released. Condition (3) holds as
  no vehicle can be removed. Condition (4) hold as
  there is a circular wait (5 points). Note you may
  also argue that the circle can be broken and
  there is no deadlock.
• There are many examples of deadlocks in
  telecommunication applications. One example is
  teleconferencing you should pick a reasonable
  one (5 points).

8
Problem 5 Memory hierachy

• Solution On average access to main memory takes
  110ns 0.1ns (3 points). On average access to
  disk takes 10.25ms 0.1ns (3 points). The
  expected time to get a word is 0.30.10.1
  0.5us (2 points). On average the CPU can access
  1sec/0.5us 2M words (2 points).

9
Problem 6 Memory Addressing

• Please see the homework 2 (problem 1) for a
  simple draw of the addressing mechanisms (5
  points). The virtual address space has 32 bits,
  but really only 30 bits are useful. See the
  follows
•  
• For the virtual address 0x83A4C234, binary 1000
  0011 1010 0100 1100 0010 0011 0100, its segment
  is 2, its page is 0x3A4C, its offset is 0x234.
  Let us assume in the page table of segment 2,
  page 0x3A4C contains 0xDDDD. Then the physical
  address is 0xDDDD234 (5 points).
•  
• For the virtual address 0x94A4D234, binary 1001
  0100 1010 0100 1101 0010 0011 0100, it is invalid
  since two useless bits are set (5 points).
  Therefore it causes overflow. Note if you ignore
  this you may still get an answer.

10
Problem 7 Replacement Algorithms

• Part one easy 5 points
• Part two new algorithms, table (5 points)
• The new algorithm is better. This example shows
  the new algorithm tries to keep the hot items
  (items 1 and 2) in the cache, and it pays off
  eventually. It does so by remembering the number
  of accessed to the items (using a count for each
  item). This algorithm is a more complicated than
  the LRU algorithm. You may also argue that the
  new algorithm may not work well for other access
  patterns. (5 points)

11
Problem 8 File Access

• Solution First the root directory / is open,
  and checked for its sub-directory ssh. Then
  ssh is open, and checked for the file
  ssh_config. Finally, ssh_config is open (5
  points). There are at least 2 disk read
  operations, one to read / and one to read ssh
  (2 points). It is possible that the system needs
  more disk reads, for example if either the
  directory / or ssh is very large and needs
  several disk pages (3 points).

12
Problem 9 File Organization
File pointer

• (1) There are 4 record pages and an index page.
  The index page contains many pairs of (phone,
  page). Each pair shows that the records with
  phone in a specific range are stored in that
  page. (5 points)
• (2) To search a record with key 617-353-8924, we
  first compare this with those in the index
  page. Then we know which record page potentially
  contains the record. Then we can read that record
  page to find the record. (4 points). Two disk I/O
  are required to get the record, one for the index
  and one for the record page. (1 point)
• (3) If there are 256000 records in the file,
  then a single index page will not work, instead
  we need more index pages (probably need up to
  dozens). One simple solution is to link these
  index pages together (5 points). But then if we
  have two many index pages, possibly we need to go
  through all of them to find a record! A better
  solution is to create a higher-level index page
  for the index pages. (It is ok if you do not
  answer in this way).

Index
64 records
64 records
64 records
64 records
13
Problem 10 Disk Scheduling

• The Elevator algorithm is better in term of
  efficiency (it finishes all jobs in 23 units of
  time while FCFS needs 29 units. FCFS is better in
  terms of fairness in that it finishes the jobs in
  order, while with Elevator, some earlier jobs may
  wait for a longer time. (5 points)

14
Introduction to Multimedia (1)

• Video On Demand (a) ADSL vs. (b) cable

15
Introduction to Multimedia (2)

• Some data rates
• multimedia, high performance I/O devices

16
Multimedia Files

• A movie may consist of several files

17
Audio Encoding (1)

• Audio Waves Converted to Digital
• electrical voltage input
• binary number as output

18
Audio Encoding (2)

• Error induced by finite sampling
• called quantization noise
• Examples of sampled sound
• telephone pulse code modulation
• audio compact disks

19
Video Encoding

• Scanning Pattern for NTSC Video and Television

20
Multimedia Operating Systems

• Multimedia job scheduling
• Disk scheduling
• Multimedia file placement and caching
• Admission control

21
The Notion of Real-Time

• A real-time process is a process which delivers
  the results of the processing in given time-span.
• The system must enforce externally-defined time
  constraints.
• Speed and efficiency are not the main
  characteristics of a real-time system.
• The playback of a video sequence is only
  acceptable when it is presented neither too
  quickly nor too slowly.
• Timing and logical dependencies among different
  related tasks, processed at the same time, also
  must be considered.
• Audio data sometimes must be synchronized with
  video data.

22
Real-time Scheduling

• To fulfill the timing requirements of continuous
  media, the operating system must use real-time
  scheduling technique.
• The scheduler must consider the entire end-to-end
  data path.
• The CPU is just one of the resources.
• Other components include main memory, storage,
  I/O devices and networks.

23
Multimedia Process Scheduling

• Periodic processes displaying a movie
• Frame rates and processing requirements may be
  different for each movie

24
Rate Monotonic Scheduling

• The users are assigned priorities such that a
  shorter fixed period between deadlines is
  associated with a higher priority.
• Used for processes which meet these conditions
• Each periodic process must complete within its
  period
• No process dependent on any other process
• Each process needs same CPU time each burst
• Any nonperiodic processes have no deadlines
• Process preemption occurs instantaneously, no
  overhead

25
Earliest Deadline First Scheduling (1)

• Real Time Scheduling algorithms
• RMS
• EDF

26
Earliest Deadline First Scheduling (2)
Another example of real-time scheduling with RMS
and EDF
27
Multimedia File System Paradigms

• Pull and Push Servers

28
Disk Scheduling for Multimedia
Stream
Order in which disk requests are processed ?

• Static Disk Scheduling
• In one round, each movie asks for one frame

29
Dynamic Disk Scheduling

• Scan-EDF algorithm
• uses deadlines cylinder numbers for scheduling

30
Multimedia Operating Systems

• Multimedia job scheduling
• Disk scheduling
• Multimedia file placement and caching
• Admission control

31
Multimedia File Placement

• How to organize a file on disks?
• How to store multiple files?
• How to store a file across multiple disks?

32
Two Alternative File Organization Strategies (1)

• Noncontiguous Movie Storage
• (a) small disk blocks (b) large disk blocks

33
Two Alternative File Organization Strategies (2)

• Trade-offs between small, large blocks
• Frame index
• heavier RAM usage during movie play
• little disk wastage
• Block index (no splitting frames over blocks)
• low RAM usage
• major disk wastage
• Block index (splitting frames over blocks
  allowed)
• low RAM usage
• no disk wastage
• extra seeks

34
Placing Multiple files on a Single Disk (1)

• Zipf's law for N20
• Squares for 20 largest cities in US
• sorted on rank order

35
Placing Multiple files on a Single Disk (2)

• Organ-pipe distribution of files on server
• most popular movie in middle of disk
• next most popular either on either side, etc.

36
Placing Files on Multiple Disks

• (a) No striping
• (b) Same striping pattern for all files
• (c) Staggered striping
• (d) Random striping

37
Multimedia File Caching

• Block Caching
• (a) Two users, same movie 10 sec out of sync (b)
  Merging two streams into one

38
Multimedia File Caching

• Most movies stored on DVD or tape
• copy to disk when needed
• results in large startup time
• keep most popular movies on disk
• Can keep first few min. of all movies on disk
• start movie from this while remainder is fetched

39
Multimedia Operating Systems

• Multimedia job scheduling
• Disk scheduling
• Multimedia file placement and caching
• Admission control

40
Admission Control

• An admission controller determines whether a new
  client can be admitted for service without
  disturbing the clients being served.
• Once the client is admitted, its requirements
  must be satisfied during the course of service.
• After a new client is admitted, the scheduler
  schedules the client of when and how it is served.

41
Admission Control
Video files
42
Admission Control Goals

• To service as many clients as possible per unit
  time (high throughput).
• To maintain a high utilization of the resources.
• To offer minimum latency for the clients.

43
Admission Criterion

• Service time is the total time spent retrieving
  media blocks of currently served clients for one
  round.
• Round duration is the minimum playback duration
  among the currently served clients for a round.
• The admission criterion is
• Service_time ? Round_duration

44
Deterministic Admission Control

• Different admission control techniques compute
  the service time differently.
• Deterministic approach assumes the worst-case
  scenarios in computing the service time.
• Service_time ? Round_duration

45
Deterministic Admission Control

• Advantages
• The continuity requirements of each client are
  not violated during the entire course of their
  playback.
• The admission control algorithm is easy to
  implement.
• Disadvantages
• The media server is underutilized since the
  average time for retrieving a block is usually
  much lower than the worst case value.
• The throughput of the system is much less than
  the peak one.

46
Statistical Admission Control

• It extrapolates the average block retrieval time
  in future rounds based upon the history of the
  average retrieval times of the most W recent
  rounds.
• It admits the client if the following criterion
  is satisfied
• Service_time ? Round_duration

47
Statistical Admission Control

• Advantages
• The server resources are better utilized. As a
  result, throughput is significantly increased.
• Disadvantages
• It does not provide absolute guarantee to the
  clients since the algorithm employs prediction.
• There are overflow rounds.
• Several techniques can be applied to distribute
  the media loss among clients.
• The algorithm is more complicated to implement.

48
Reading

• Chapter 7
• Section 7.1, 7.2, 7.4, 7.6, 7.8

49
Next Week

• Multimedia communication protocols
• Homework assignment 3