A New Retrieval Strategy A New Retrieval Strategy for P2P Video-On-Demand for P2P Video-On-Demand Systems Systems Presented By… Ashwini Ramesh More Mounika Eluri CS 696 – Advanced Distributed
A New Retrieval Strategy A New Retrieval Strategy for P2P Video-On-Demand for P2P Video-On-Demand
SystemsSystems
Presented By… Ashwini
Ramesh More Mounika Eluri
CS 696 – Advanced Distributed SystemSan Diego State University
AGENDA
2
INTRODUCTIONVoD (Video on Demand) - allows users to
select and watch/listen to video content whenever they want.
Necessity to provide instantaneous response to end-users.
Delivering the media content over the network with best response time has been a popular topic of many discussions.
Our objective is to design a retrieval strategy to achieve minimum response time and maximize the overall throughput of the system.
3
DRAWBACK OF LEAST LOAD FIRST
It selects a serving peer having the least load for delivering the media content.
Since only one peer is responsible for servicing the request, it takes more time to respond to the request thereby affecting response time.
4
MOTIVATIONLeast Load First strategy is time
consuming.Can we develop an algorithm which can
reduce the mean response time ?We propose an algorithm called
CollaborativeRetrieval (CoRe) algorithm which aims in minimizing the mean response time.
5
CORE MODEL
10 40
3020
C
B
A
D
X
6
S
CORE MODEL
10
30%
40
3020
40%
20%
10%
C
B
A
D
X
7
S
Collaborative Retrieval
CORE ALGORITHMInput: Batch of movie requests, list of available peers, list of movie replicas
distributed across multiple peers.Output: response time for each request 1. for each request ri do 2. size = getSize(ri) 3. Get list of available peers containing the movie and store in list Lp
4. Total = count (Lp) 5. for each peer pi in list Lp do 6. Set distance with respect to the request source 7. end for 8. Sort the list Lp according to the distance factor in ascending order 9. for each peer pi in list Lp do 10. Calculate the cost, cost [pi] = distance [pi]/Total 11. Request_service_time = size * cost [pi]/transfer_rate (31Kbps
assumed) 12. end for 13. Record start time and end time for request ri 14. end for
8
EXPERIMENTAL PARAMETERS
Parameter ValuesNumber of requests 2000-15000Number of peers 100Number of movies
500
Skew 50-50, 60-40, 70-30Aggregate access rate (1/s)
50, 100, 150, 200, 250, 300
9
10
MEAN RESPONSE TIME (SKEW 70-30)
MEAN RESPONSE TIME (SKEW 60-40)
11
12
MEAN RESPONSE TIME (SKEW 50-50)
RESPONSE IMPROVEMENT (SKEW 70-30)
13
RESPONSE IMPROVEMENT (SKEW 60-40)
14
RESPONSE IMPROVEMENT (SKEW 50-50)
15
CONCLUSIONWe proposed an efficient CoRe strategy for
retrieving the videos. Our experimental results showed that CoRe
performs significantly better than existing Least Load First algorithm even in the case of heavy workload.
Simulations performed for skew distribution of 70-30 showed that CoRe algorithm achieved the maximum improvement of 36 percent over Least Load First. 16
FUTURE WORK Further studies in this research can be
performed by taking into consideration the issues like,
Data corruptionPeer or network failure and recovery
17
Thank You !!!
18