Download slides - Network and Systems Laboratory

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Net bias wikipedia , lookup

Remote Desktop Services wikipedia , lookup

Distributed firewall wikipedia , lookup

Lag wikipedia , lookup

Cracking of wireless networks wikipedia , lookup

Network tap wikipedia , lookup

IEEE 802.1aq wikipedia , lookup

CAN bus wikipedia , lookup

Airborne Networking wikipedia , lookup

Deep packet inspection wikipedia , lookup

List of wireless community networks by region wikipedia , lookup

Peer-to-peer wikipedia , lookup

Transcript
An Experimental Study of the
Skype Peer-to-Peer VoIP System
Saikat Guha, Cornell University
Neil DasWani, Google
Ravi Jain, Google
IPTPS’06
Presenter: Te-Yuan
What do they want to know?
 What makes Skype so successful?
 Compare with
 File-sharing P2P network
 By Observing Skype’s
 User behavior
 Node Session Time
 Overlay Network Traffic
 SuperNode overlay network
 Overall utilization & resource consumption
Skype
 Three Services
 two-way audio streams & conference call up to 4
users
 Instant Message
 file-transfer
 Structure
 Alike KaZaA
– SuperNode-based
 Ordinary Node (ON)
 Super Node (SN)
Outline - Experiments
Expt. 1: Basic operation
Expt. 2: Promotion to supernode
Expt. 3: Supernode network activity
Expt. 4: Supernode and client
population
 Expt. 5: Supernode presence




Expt. 1: Basic operation
 To Answer: How do two Skype clients
connect to each other?
 Normally,
 ON send control traffic through SN-p2p
 Including
 Availability information
 Instant messages
 Request for VoIP & File-transfer
 What if ON is behind NAT/Firewall?
Expt. 1: Basic operation – Cont.
 NAT Traversal in Skype:




Level
Level
Level
Level
0:
1:
2:
3:
Initiator NAT’ed
Recipient NAT'ed
Both NAT'ed (well-behaved NATs)
Both NAT'ed
Expt. 1: Basic operation – Cont.
 Level 0: Initiator NAT’ed
Expt. 1: Basic operation – Cont.
 Level 1: Recipient NAT’ed
Expt. 1: Basic operation – Cont.
 Level 2: Both NAT'ed (well-behaved
NATs)
Expt. 1: Basic operation – Cont.
 Level 3: Both NAT'ed
Expt. 1: Basic operation – Cont.
Level 0
Level 1
Level 2
Level 3
Outline - Experiments
Expt. 1: Basic operation
Expt. 2: Promotion to supernode
Expt. 3: Supernode network activity
Expt. 4: Supernode and client
population
 Expt. 5: Supernode presence




Expt. 2: Promotion to supernode
 To Answer: What kind of node will be
promote to SN?
 Setup several Skype clients
 One behind a saturated network uplink
 One behind a NAT
 One with a 10 Mbps connection & public IP
 Key to be SN
 plenty of spare bandwidth
 publicly reachable
Outline - Experiments
 Expt. 1: Basic operation
 Expt. 2: Promotion to supernode
 Expt. 3: Supernode network activity
 Expt. 4: Supernode and client
population
 Expt. 5: Supernode presence
Expt. 3: Supernode network
activity
 Goal: To observe the network traffic
of a Skype supernode
 Duration: 135 days (Sep. 1, 2005 to
Jan. 14, 2006)
 Data captured: 13GB with ethereal
Expt. 4: Supernode and client
population
 Goal: Collect SN & client IP/port
 Duration:2005/7/25 – 2005/10/12
 Result:
 Crawl 150K SN
 Collect 250K SN info
Expt. 4: Supernode and client
population
Connect to a SN
A list of SN
Save the list
Connect to a SN from the list
A list of SN
Expt. 4: Supernode and client
population
 Collect client info
 Collect the number
reported by skype
client
Expt. 5: Supernode presence
 Goal: how many SN online at a give
time
 Flow
 Randomly Select 6000 SN
- from the list collected by expt. 4
 Send “application-layer Ping”
 Repeat every 30 mins for a month
Expt. 5: Supernode presence Cont
Num. of SuperNode
is more Stable
Weekend
diurnal behavior of
SN
Expt. 5: Supernode presence Cont
 Geographic Distribution of Active SuperNodes
15-25%
20-25%
45--60%
peak at 11am UTC
(Europe mid-day)
Expt. 5: Supernode presence Cont
 SuperNode Session Time
Median is 5.5h
Expt. 5: Supernode presence Cont
 Fraction of supernodes joining or departing
Node arrival
concentrated
toward morning
Skype usage is correlated with working hours
Different from P2P file-sharing
Node departure
concentrated
toward evening
Expt. 5: Supernode presence Cont
 Node Arrival dependent
on Time
 Not Poisson or Uniform
process
 Poisson process with
varying hourly rate
Node arrival
concentrated
toward morning
Node departure
concentrated
toward evening
VoIP in Skype: Preliminary
Observation
 SuperNode Traffic
90.4%SN no need
to relay VoIP
traffic
VoIP in Skype: Preliminary
Observation
 VoIP Relayed Session Arrival Behavior
 Inter arrival time of Relayed VoIP/File sessions may
be Poisson
VoIP in Skype: Preliminary
Observation
 VoIP Session Length Behavior
Skype:
Median: 2m50s
Average: 12m53s
Longest: 3h 26s
Traditional:
Average: 3m
Fraudulent:
Average: 9m
VoIP in Skype: Preliminary
Observation
 File-transfer sizes
File size:
Median: 346kB
Conclusion
 First measurement study of Skype VoIP
system
 Skype differs significantly from file-sharing
P2P
 User Behavior
 Diurnal & Work-week
 Calls are significantly longer
 File transferred are significantly smaller
 SuperNode of Skype
 Consume little bandwidth
 Relatively stable