Modeling and Analysis of Reliable Peer-to-Peer Storage Systems

Julian Monteiro 1
1 MASCOTTE - Algorithms, simulation, combinatorics and optimization for telecommunications
CRISAM - Inria Sophia Antipolis - Méditerranée , Laboratoire I3S - COMRED - COMmunications, Réseaux, systèmes Embarqués et Distribués
Abstract : Large scale peer-to-peer systems are foreseen as a way to provide highly reliable data storage at low cost. To ensure high durability and high resilience over a long period of time the system must add redundancy to the original data. It is well-known that erasure coding is a space efficient solution to obtain a high degree of fault-tolerance by distributing encoded fragments into different peers of the network. Therefore, a repair mechanism needs to cope with the dynamic and unreliable behavior of peers by continuously reconstructing the missing redundancy. Consequently, the system depends on many parameters that need to be well tuned, such as the redundancy factor, the placement policies, and the frequency of data repair. These parameters impact the amount of resources, such as the bandwidth usage and the storage space overhead that are required to achieve a desired level of reliability, i.e., probability of losing data. This thesis aims at providing tools to analyze and predict the performance of general large scale data storage systems. We use these tools to analyze the impact of different choices of system design on different performance metrics. For instance, the bandwidth consumption, the storage space overhead, and the probability of data loss should be as small as possible. Different techniques are studied and applied. First, we describe a simple Markov chain model that harnesses the dynamics of a storage system under the effects of peer failures and of data repair. Then we provide closed-form formulas that give good approximations of the model. These formulas allow us to understand the interactions between the system parameters. Indeed, a lazy repair mechanism is studied and we describe how to tune the system parameters to obtain an efficient utilization of bandwidth. We confirm by comparing to simulations that this model gives correct approximations of the system average behavior, but does not capture its variations over time. We then propose a new stochastic model based on a fluid approximation that indeed captures the deviations around the mean behavior. These variations are most of the time neglected by previous works, despite being very important to correctly allocate the system resources. We additionally study several other aspects of a distributed storage system: we propose queuing models to calculate the repair time distribution under limited bandwidth scenarios; we discuss the trade-offs of a Hybrid coding (mixing erasure codes and replication); and finally we study the impact of different ways to distribute data fragments among peers, i.e., placement strategies.
Complete list of metadatas

Cited literature [152 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00545724
Contributor : Julian Monteiro <>
Submitted on : Saturday, December 11, 2010 - 3:19:29 PM
Last modification on : Monday, April 29, 2019 - 4:13:03 PM
Long-term archiving on : Saturday, December 3, 2016 - 12:03:27 AM

Identifiers

  • HAL Id : tel-00545724, version 1

Collections

Citation

Julian Monteiro. Modeling and Analysis of Reliable Peer-to-Peer Storage Systems. Networking and Internet Architecture [cs.NI]. Université Nice Sophia Antipolis, 2010. English. ⟨tel-00545724⟩

Share

Metrics

Record views

600

Files downloads

1322