## Impact of QoS on Replica Placement in Tree Networks (2006)

### Cached

### Download Links

- [graal.ens-lyon.fr]
- [hal.inria.fr]
- [hal.archives-ouvertes.fr]
- DBLP

### Other Repositories/Bibliography

Citations: | 5 - 2 self |

### BibTeX

@TECHREPORT{Benoit06impactof,

author = {Anne Benoit and Veronika Rehn and Yves Robert},

title = {Impact of QoS on Replica Placement in Tree Networks},

institution = {},

year = {2006}

}

### OpenURL

### Abstract

This paper discusses and compares several policies to place replicas in tree networks, subject to server capacity and QoS constraints. The client requests are known beforehand, while the number and location of the servers are to be determined. We study three strategies. The first two strategies assign each client to a unique server while the third allows requests of a client to be processed by multiple servers. The main contribution of this paper is to assess the impact of QoS constraints on the total replication cost. In this paper, we establish the NP-completeness of the problem on homogeneous networks when the requests of a given client can be processed by multiple servers. We provide several efficient polynomial heuristic algorithms for NP-complete instances of the problem. These heuristics are compared to the optimal solution provided by the formulation of the problem in terms of the solution of an integer linear program. 1

### Citations

11403 |
Computers and Intractability: A Guide to the Theory of NP-Completeness
- Garey, Johnson
- 1979
(Show Context)
Citation Context ... 1 2 1 1 n2m+1 n2m+2 n2m+3 n1 n2 n2m a1 b1 a2 b2 a2m b2m 1 1 n3m−1 Figure 2: The platform used in the reduction for Theorem 2. To establish the completeness, we use a reduction from 2-PARTITION-EQUAL =-=[5]-=-. We consider an instance I1 of 2-PARTITION-EQUAL: given 2m positive integers a1, a2, . . . , a2m, does there exist a subset I ⊂ {1, . . . , 2m} of cardinal m such that � i∈I ai = � i/∈I ai. Let S = �... |

1529 |
Theory of Linear and Integer Programming
- Schrijver
- 1986
(Show Context)
Citation Context ...nts and solve the linear program assuming that all variables take rational values. The optimal solution of the relaxed program can be obtained in polynomial time (in theory using the ellipsoid method =-=[9]-=-, in practice using standard software packages [2, 6]), and the value of its objective function provides an absolute lower bound on the cost of any valid (integer) solution. For all practical values o... |

74 | Dynamic Replica Placement for Scalable Content Delivery
- Chen, Katz, et al.
- 2002
(Show Context)
Citation Context ...communication links. A heuristic algorithm is proposed and compared to the results of Tang and Xu [11]. Another approach, this time for dynamic content distribution systems, is proposed by Chen et al =-=[3]-=-. They present a replica placement protocol to build a dissemination tree matching QoS and server capacity constraints. Their work focuses on Web content distribution built on top of peer-topeer locat... |

49 | Optimal allocation of electronic content
- Cidon, Kutten, et al.
- 2001
(Show Context)
Citation Context ...nstead consider average system performance such as total communication cost or total accessing cost. Please refer to [1] for a detailed description of related work with no QoS contraints. Cidon et al =-=[4]-=- studied an instance of Replica Placement with multiple objects, where all requests of a client are served by the closest replica (Closest policy). In this work, the objective function integrates a co... |

38 | A framework for evaluating replica placement algorithms
- Karlsson, Karamanolis, et al.
- 2002
(Show Context)
Citation Context ...erver and link capacities). We derive a formulation for each of the three server access policies, namely Closest, Upwards and Multiple. This is an important extension to a previous formulation due to =-=[7]-=-. While there is no efficient algorithm to solve integer linear programs (unless P=NP), this formulation is extremely useful as it leads to an absolute lower bound: we solve the integer linear program... |

18 | M.J.: QoS-aware replica placement for content distribution
- Tang, Xu
- 2005
(Show Context)
Citation Context ... their capacity is bounded. Both [4, 8] use a dynamic programming algorithm to find the optimal solution. Some of the first authors to introduce actual QoS constraints in the problem were Tang and Xu =-=[11]-=-. In their approach, the QoS corresponds to the latency requirements of each client. Different access policies are considered. First, a replica-aware policy in a general graph with heterogeneous nodes... |

17 |
Optimal placement of replicas in data grid environments with locality assurance
- Liu, Lin, et al.
- 2006
(Show Context)
Citation Context ... a given QoS for each client. They target fully homogeneous platforms since there are no server capacity constraints in their approach. A similar instance of the problem has been studied by Liu et al =-=[8]-=-, adding a QoS in terms of a range limit (QoS=distance), and whose objective is to minimize the number of replicas. In this latter approach, the servers are homogeneous, and their capacity is bounded.... |

7 | Strategies for Replica Placement in Tree Networks
- Benoit, Rehn, et al.
- 2007
(Show Context)
Citation Context ...plica Placement optimization problem. Most of the papers do not deal with QoS but instead consider average system performance such as total communication cost or total accessing cost. Please refer to =-=[1]-=- for a detailed description of related work with no QoS contraints. Cidon et al [4] studied an instance of Replica Placement with multiple objects, where all requests of a client are served by the clo... |

6 | J.J.: A QoS-aware heuristic algorithm for replica placement
- Wang, Liu, et al.
(Show Context)
Citation Context ...are (replicablind policy), the graph is simplified to a tree (fixed routing scheme) with the Closest policy, and in this case again it is possible to find an optimal dynamic programming algorithm. In =-=[12]-=-, Wang et al deal with the QoS aware replica placement problem on grid systems. In their general graph model, QoS is a parameter of communication cost. Their research includes heterogeneous nodes and ... |