Review of a research paper on Skype

Review of a research paper on Skype

• Ashwin Sampath

Outline

• Introduction

• Motivations of research

• Goals of research

• Adopted techniques

• Analysis

• Conclusions

Introduction

• An Analysis of the Skype Peer to Peer Internet Telephony Protocol

• - Salman A Baset & Henning Schulzrinne

Motivation

• Verify the functionality and quality claims made by Skype.

Goals

• Analyze the basic functionalities of Skype

• Analyze the network traffic generated by Skype under different network configurations

Techniques Adopted

• Reverse-engineering based approach.

• Run Skype clients on windows machines.

• Three different network setups with machines behind NAT and firewall.

• Analyze network traffic using Ethereal.

• Tune network bandwidth using NetPeeker.

Background of Skype

• P2P voip client supporting voice and text based conversation, buddy lists etc.

• Based on Kazaa.

• Overlay p2p network consisting of ordinary and Super Nodes (SN).

• Ordinary node connects to network through a Super Node.

• Each user registers with a central server.

• User information propagated in a decentralized fashion.

• Uses a variant of STUN to identify the type of NAT and firewall.

Background of Skype continued..

• Skype client listens on configured TCP and UDP ports.

• Each client maintains a list of super nodes in the Host Cache.

• Buddy list is local to a machine.

• All Skype messages are encrypted.

Key observations

• Initial startup and login sequences are different from subsequent sequences.

• Usage of TCP port 80 enables client to reach super node even through firewalls.

• Skype login server might be fixed but not hardcoded.

• Identity of “Bootstrap” nodes hardcoded.

• Client needs to maintain a constant TCP connection with atleast one super node.

• Skype client continuously discovers and builds the list of Skype nodes.

Key observations continued..

• When the client is behind a UDP-restricting firewall, the login takes significantly more time.

• Search results are cached at intermediate nodes.

• Call signaling is done over TCP, messages are preferably transported over UDP.

• In the presence of NAT or firewalls, calls between caller and callee are routed by an intermediate node.

• Silence suppression is not supported in order to maintain UDP bindings and TCP congestion window size.

• Reasonable call quality requires atleast 2kbps.

• During conferencing traffic always passes through one node which can become a single point of failure.

Strong points

• “Real-time” examination of working of Skype.

Weak Points

• Observations are inconclusive.

• eg : Caching of search results.

• Organization of figures is bad.

• Figures contradictory to text explanation

• eg : Figure 11 shows UDP message exchange and media transport over UDP.

• Poor conclusions.

Results

• Growing popularity of Skype because of better voice quality, ability to work behind firewalls and NAT and ease of use.