peer-to-peer based multimedia distribution service
DESCRIPTION
Peer-to-Peer Based Multimedia Distribution Service. Zhe Xiang, Qian Zhang , Wenwu Zhu , Zhensheng Zhang , and Ya-Qin Zhang IEEE TRANSACTIONS ON MULTIMEDIA, April, 2004. Outline. Introduction P2p based multimedia distribution service architecture Design of the system - PowerPoint PPT PresentationTRANSCRIPT
Peer-to-Peer Based Multimedia Distribution Service
Zhe Xiang, Qian Zhang, Wenwu Zhu, Zhensheng Zhang, and Ya-Qin Zhang
IEEE TRANSACTIONS ON MULTIMEDIA, April, 2004
Outline
Introduction P2p based multimedia distribution service
architecture Design of the system
Topology-aware overlay Replication strategies
Simulation results Conclusion
Characteristics of multimedia distribution service in p2p system
A peer shares data with other peers A peer searches data from other peers Dedicated server is not needed Replica is allowed
P2p based multimedia distribution service architecture
How does a new node join
Content Delivery
Node A of group G requests for content C If C is in group G: transfer directly If C is not in group G: transfer C to G first
Replication G may make a replication for delay concern
(inter-group replication strategy) A may make a replication for robust concern
(intra-group replication strategy)
Inter-group replication strategy
Peer pj in group Gr requests content ci
Find nearest group Gp that contains ci
Calculate avg. distance d between neighbors If d(Gp,Gr)<d: end If (free space>size of ci): replicate ci in local stor
age, end Else: remove stored contents according to avail
ability (or maybe LRU, LFU,…etc.) Replicate content ci into local storage
Problem formulation
Seed replication matrix X={xij}: xij=1 if Gj holds seed for content ci
xij=0 otherwise
Access matrix Y={yij} yij=k: k≠j, d(Gj,Gk)=min{d(Gj,Gs), xis=1}
Inter-group replication:
M
i
N
jyji jiGGdr
1 1
)),((,
subject to
M
ijijji NjxSx
1,, .,...,2,1},1,0{,)(
minimize
Theoretical analysis
Only consider total capacity:
M
i
N
jji SS
1 1
Avg. distance between neighbors:
2
12
)(
ii
i LL
cd
2
1
)()(
iii Lcdcd
Popularity factor
Assume popularity follows a Zipf distribution Suppose M contents are ordered in a desce
nding order of popularity ri:
Normally, skew factor(α) is chosen as 0.271 The authors vary skew factor from 0.1 to 0.5
)1(
)1(
i
iri
Intra-group replication strategy
Peer pj requests content cr within its group Calculate A(cr) Calculate A’(ci) for any content ci in pj’s storage Create list {c1,c2…ck} in a descending order of A’(ci) If (FreeStorage+total size of contents in list)<sr: end Find j in list such that
Delete contents c1,…cj
Replicate content cr
Problem formulation
N
jjiji xcA
1, )1(1)(
Replica replication matrix X{xij} xij=1 if per pj holds a replica of content ci
Xij=0 otherwise
Availability A(ci) of content ci:
Simulation methodology
Edge longitude = 300 ms Group number: 200 Latency:
Small if in the same group According to distance if in different group
# of hops increases linearly to distance Bandwidth ranges from 800 kbps to 1.4 Mbps 10000 video chips, each is encoded in 1.28 Mbps
bit rate length of video clips: 3~5 min
Simulation results
Simulation results (cont’)
Simulation results (cont’)
Simulation results (cont’)
Conclusion
Proposed a framework for multimedia distribution based on p2p network Hosts are self-organized into groups Proposed heuristic method to make replica in
the system