Distributed Multimedia Systems

25
1 Distributed Multimedia Systems Resource Management Stream Adaptation Case Study The Tiger Video file server

description

Distributed Multimedia Systems. Resource Management Stream Adaptation Case Study The Tiger Video file server. Resource Management. Resource Scheduling. - PowerPoint PPT Presentation

Transcript of Distributed Multimedia Systems

Page 1: Distributed Multimedia Systems

1

Distributed Multimedia Systems

Resource Management Stream Adaptation Case Study The Tiger Video file server

Page 2: Distributed Multimedia Systems

2

Resource Management Resource SchedulingTo provide Quality of Service (Qos) to an application not only system must have sufficient resource (performance), it also needs to make these resource available to an application when they are needed (scheduling).

Page 3: Distributed Multimedia Systems

3

Resource Scheduling

Fair Scheduling Real-time Scheduling

Page 4: Distributed Multimedia Systems

4

Fair Scheduling If several streams compete for a same

resource it is necessary to consider fairness and to prevent ill behaved streams taking too much bandwidth.

Round robin method is used on bit by bit basis, which provides more fairness with respect to varying packet sizes and arrival times.

Page 5: Distributed Multimedia Systems

5

Real-time Scheduling The Scheduling algorithms assigns

CPU time slots to a set of processes in a manner that ensures that they complete their tasks on time.

Earliest- deadline first (EDF).

Page 6: Distributed Multimedia Systems

6

Stream Adaptation Adjustment in Qos Droping a piece of information

(audio) Dropouts in video stream MPEG We Use scaling methods for

dropouts.For video files we use the

following scaling methods or combination

of it.

Page 7: Distributed Multimedia Systems

7

Video Scaling methods Temporal Scaling Spatial Scaling Frequency Scaling Amplitudinal Scaling Color space Scaling

Page 8: Distributed Multimedia Systems

8

Scaling Temporal Scaling reduces the resolution of video

stream in the time domain by decreasing the number of video frames transmitted with in a interval.

Page 9: Distributed Multimedia Systems

9

Scaling Spatial Scaling reduces the number of pixels of

each image in a video stream.

Frequency Scaling modifies compressed algorithm

applied to a image.

Page 10: Distributed Multimedia Systems

10

Scaling Amplitudinal Scaling Reduces the color depth of each

image pixel Color space Scaling Reduces the number of entities in

the color space color -> gray scale

Page 11: Distributed Multimedia Systems

11

Filtering Scaling modifies the stream of source

it is not suitable for applications that involve several receivers: if bottleneck occurs on the route of one target, This target sends Scale-down message to the source and all targets receive the degraded quality,although some do not require.

Page 12: Distributed Multimedia Systems

12

Filtering Filtering is a method that provides

the best possible quality of service to each target applying.

Filtering requires that a stream be partitioned into a set of hierarchical sub streams, each adding a higher level of quality.

Page 13: Distributed Multimedia Systems

13

Filtering

Source

Targets

Page 14: Distributed Multimedia Systems

14

Case Study A video storage system that

supplies multiple real time video streams simultaneously is seen as an important system component to support consumer-oriented multimedia applications.

Tiger video file server developed by Microsoft research Labs.

Page 15: Distributed Multimedia Systems

15

Case Study : The Tiger video file server Design goals Architecture Storage organization Distributed schedule Network support

Page 16: Distributed Multimedia Systems

16

Design Goals Video on demand for a large

number of users Quality of service Scalable and distributed Low cost hardware

Page 17: Distributed Multimedia Systems

17

Architecture The Cub Computers are identical

PC’s with same number of standard Hard disk drives attached to each. They are equipped with ethernet and ATM network cards. The Controller is another PC it handles the client requests and manages the work schedule of the curb.

Page 18: Distributed Multimedia Systems

18

ArchitectureController

Cub 0 Cub 1 Cub 2 Cub n

ATM Switching network

Video Distribution to clients

start/Stop requestsFrom clients

Low-bandwidth networkController

0…n+11…n+2 2…n+3 n…2n+1

Page 19: Distributed Multimedia Systems

19

Storage organization Video data is a large file in order to

share the load its distributed among the disks attached to the cubs.

A movie is divided into blocks ( 1sec -> 0.5MB so a 2 hr movie has app. 7000) Movie can start on any disk whenever highest numbered disk is reached, the movie is wrapped around so that next block disk 0 is in process.

Page 20: Distributed Multimedia Systems

20

Distributed Schedule Scheduling workload for the cubs. Schedule is organized as list of

slots Each slot rep. Work must be done

to play one block of movie read it from relevant disk transfer it to ATM network.

Page 21: Distributed Multimedia Systems

21

Network support The blocks of each movie are simply

passed to the ATM network by the cubs that hold them, together with the address of the relevant client.

Client needs sufficient buffer storage to hold two buffer locations 1 is playing the video and other is arriving from the network.

Page 22: Distributed Multimedia Systems

22

Problem 1 Outline the design of a QoS manager

to enable desktop Computers connected by an ATM network to support several concurrent multimedia applications. Define an API for your QoS manager, giving the main operations with their parameters and results.

Page 23: Distributed Multimedia Systems

23

Problem 2 In order to specify the resource

requirements software components that process multimedia data, we need estimates for their processing loads. How can this information can be obtained without undue effort?

Page 24: Distributed Multimedia Systems

24

Problem 3 The Tiger schedule is potentially a

large data structure that changes frequently, but each cub needs an up-to-date representation of the portions it is currently handling. Suggest a mechanism for the distribution of the schedule to the cubs.

Page 25: Distributed Multimedia Systems

25

Problem 4 When Tiger is operating with a

failed disk or cub, secondary data blocks are used in place of missing primaries. Secondary blocks are n times smaller than primaries ( where n is the decluster factor), how does the system accommodate this variability in block size?