Quality of service routing with performance considerations in peer-to-peer systems

Rakaseljar, S 2007, Quality of service routing with performance considerations in peer-to-peer systems, Doctor of Philosophy (PhD), Computer Science and Information Technology, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Rajasekhar.pdf Thesis Click to show the corresponding preview/stream application/pdf;... 1.26MB
Title Quality of service routing with performance considerations in peer-to-peer systems
Author(s) Rakaseljar, S
Year 2007
Abstract The dream of having information while on the move anywhere anytime, is a reality. The challenging problems of sharing information must be overcome. First, due to the limited or absence of quality of service (QoS) in present day architectures, a very reliable architecture is needed to integrate mobile devices in fixed and mobile P2P networks. Second, efficient search technique is important to locate the required information. Third, strong and robust routing schemes are required to route information between these peer devices. Fourth, the load on peers has to be distributed to lightly loaded peers to achieve better system performance.

We examine routing schemes and load sharing principles in the existing P2P architectures. The focus is on improving the architecture and providing routing schemes to efficiently use bandwidth in the P2P network. Load sharing schemes to enhance the overall performance and reliability of P2P systems are also addressed.

Extended Super-Peer based file sharing Architecture (ESPA) for mobile and fixed peer devices is essential since the existing architectures lack QoS support and seamless integration of peer devices. We investigate routing schemes for ESPA architecture. A Static Path Capacity to Hop QoS (SPCHQ) routing is proposed to demonstrate the selection of an efficient path with more available bandwidth.

We look at inaccuracies in the link state information during routing. Since, the available bandwidth changes dynamically, information becomes out-of-date. We define a probabilistic metric and propose two algorithms: Probabilistic Path Capacity (PPC) and Probabilistic Path Capacity to Hop (PPCH) that capture the dynamic nature of the available link state information and route data files in P2P networks. We also propose probabilistic QoS path selection with re-routing when congestion occurs. Since the probability of finding a good path is very much dependent on the accuracy of the available bandwidth or delay on the links in the P2P network, we propose a regression model to obtain the frequency of the occurrences of paths.


To minimise the load on super-peers, load sharing techniques such as Periodic Push based Replication (PPR) and On-Demand Replication methods are proposed since the presence of non-uniform data distribution and dynamic network conditions cause very heavy load on the ESPA. In PPR, data is pushed by a super-peer to other super-peers based only on an access frequency threshold, whereas in ODR, data can be pulled by any other super-peers at any given time. Some of the super-peers may become unavailable due to heavy traffic load. These replication schemes ignore this, and do not indicate to how many super-peers data items need to be replicated. Therefore, enhanced approaches such as Reliable Periodic Push based Replication (RPPR) and Reliable Periodic Rejuvenation based Replication (RPRR) algorithms are proposed to enhance content availability and overall system reliability.

Detailed analytical models and experiments of the proposed ESPA demonstrate that it not only reduces file download times by exploiting efficient QoS path selection algorithms, but also significantly improves the overall system reliability by deploying advanced data replication strategies.

Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Computer Science and Information Technology
Keyword(s) Super-peer
P2P architecture
QoS
Load sharing
Replication
Reliability
Versions
Version Filter Type
Access Statistics: 152 Abstract Views, 104 File Downloads  -  Detailed Statistics
Created: Wed, 14 Dec 2016, 13:38:12 EST by Denise Paciocco
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us