## Replicated Server Placement with QoS constraints (2006)

### Cached

### Download Links

- [genesis.ee.auth.gr]
- [genesis.ee.auth.gr]
- DBLP

### Other Repositories/Bibliography

Venue: | PUBLICATIONS IN CONFERENCES AND WORKSHOPS |

Citations: | 7 - 0 self |

### BibTeX

@INPROCEEDINGS{Rodolakis06replicatedserver,

author = {Georgios Rodolakis and Stavroula Siachalou and Leonidas Georgiadis},

title = {Replicated Server Placement with QoS constraints},

booktitle = { PUBLICATIONS IN CONFERENCES AND WORKSHOPS},

year = {2006},

pages = {2006},

publisher = {}

}

### OpenURL

### Abstract

The problem of placing replicated servers with QoS constraints is considered. Each server site may consist of multiple server types with varying capacities and each site can be placed in any location among those belonging to a given set. Each client can de served by more than one locationsaslongastherequestround-tripdelaysatisfies predetermined upper bounds. Our main focus is to minimize the cost of using the servers and utilizing the link bandwidth, while serving requests according to their delay constraint. This is an NP-hard problem. A pseudopolynomial and a polynomial algorithm that provide guaranteed approximation factors with respect to the optimal for the problem at hand are presented.

### Citations

10886 |
Computers and Intractability: A Guide to the Theory of NP-Completeness
- Garey, Johnson
- 1979
(Show Context)
Citation Context ...ems associated to Problems 1 and 2 are NP-hard. Indeed, when β =0, the associated decision problem to Problem 1 is reduced to the Shortest Weight-Constrained Path problem which is known to be NP-hard =-=[10]-=-. Also, the associated decision problem to Problem 2 amounts to the Unbounded Knapsack Problem which is NP-hard [11]. However for both problems pseudopolynomial algorithms exist (see Section 3) and, a... |

968 |
Computational Geometry : Algorithms and Applications
- Berg, Kreveld, et al.
- 1997
(Show Context)
Citation Context ... 0 = W +2 points in the set © (y,f(y)), y=0, 1,...,W ª ∪{(0, 0), (W, 0)} and the upper hull of the points in this set can be constructed (i.e., its break points can be determined) in time W 0 log W 0 =-=[25]-=-. However, in our case W = |V | gmax and hence the straightforward construction of the upper hull requires pseudopolynomial construction time. To address the latter problem, we construct first a non-d... |

636 | The LEDA Platform of Combinatorial and Geometric Computing
- Mehlhorn, Näher
- 1999
(Show Context)
Citation Context ...es using two different routing methods. To generate random networks a number |V | of nodes and a number |E| = γ |N| of edges, γ>1is chosen. Weusethegraphgeneratorrandom graph( ) from the LEDA package =-=[26]-=-. A random edge is generated by selecting a random element from a candidate set C defined as follows. C is initialized 16sAverage Total Cost 270000 220000 170000 120000 255795 257547 179854 198303 Ran... |

319 | Approximation algorithms for metric facility location and k-median problems using the primal-dual schema and Lagrangian relaxation
- Jain, Vazirani
(Show Context)
Citation Context ...roximation algorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], [13], [14], =-=[15]-=-, [16], [17], [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold in ou... |

288 | On the placement of Web server replicas
- Qiu, Padmanabhan, et al.
- 2001
(Show Context)
Citation Context ...b site. The problem was reduced to the placement of proxies in a set of trees whose root nodes are the server replicas. Jia et al [4] took the read and update operations into consideration. Qiu et al =-=[5]-=- also assumed a restricted number of replicas and no restriction of the number of requests served by each replica. A client could be served by a single replica and the cost for placing a replica was a... |

243 |
Knapsack problems
- Kellerer, Pferschy, et al.
- 2004
(Show Context)
Citation Context ...educed to the Shortest Weight-Constrained Path problem which is known to be NP-hard [10]. Also, the associated decision problem to Problem 2 amounts to the Unbounded Knapsack Problem which is NP-hard =-=[11]-=-. However for both problems pseudopolynomial algorithms exist (see Section 3) and, as will be discussed in Section 4, fully polynomial time approximation algorithms can be developed. Regarding Problem... |

242 | Replication strategies in unstructured peer-to-peer networks
- Cohen, Shenker
(Show Context)
Citation Context ...study on geographical replication of dynamic Web contents with the objective of minimizing the consistency management costs in terms of update transfers and 2sobject reconstruction. Cohen and Shenker =-=[13]-=- defined replication strategies in decentralized unstructured systems. They assumed each node had capacity ρ, which was the number of copies the node could hold and R was the total capacity of the sys... |

229 | Local search heuristic for k-median and facility location problems
- Arya, Garg, et al.
- 2001
(Show Context)
Citation Context ...me approximation algorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], [13], =-=[14]-=-, [15], [16], [17], [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold... |

160 |
Approximation schemes for the restricted shortest path problem
- Hassin
- 1992
(Show Context)
Citation Context ...ial Algorithms for Problem 1 and 2 A fully polynomial time approximation algorithm for the problem of finding the minimum constrained path from a source to a given destination was developed by Hassin =-=[23]-=-. An improvement of this algorithm was presented in [24]. The latter algorithm can be used as a basis for the development of a fully polynomial time approximation algorithm for Problem 1. We skip the ... |

146 | Analysis of a local search heuristic for facility location problems
- Korupolu, Plaxton, et al.
(Show Context)
Citation Context ...ial time approximation algorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], =-=[13]-=-, [14], [15], [16], [17], [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does no... |

122 | Improved approximation algorithms for uncapaciateted facility location
- Chudak
- 1998
(Show Context)
Citation Context ...lgorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], [13], [14], [15], [16], =-=[17]-=-, [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold in our case. In t... |

118 | The Cache Location Problem
- Krishnan, Raz, et al.
- 2000
(Show Context)
Citation Context ...ating servers at appropriately selected sites. The problem of maximizing the performance and minimizing the cost of a computing system has been addressed in the past in several papers. Krishnan et al =-=[1]-=- developed polynomial optimal solutions to place a given number of servers in a tree network to minimize the average retrieval cost 1sof all clients. Li et al [2] investigated the placement of a limit... |

108 | BRITE: Universal topology generation from a user’s perspective
- Medina, Lakhina, et al.
- 2001
(Show Context)
Citation Context ...e i ∈ V toasingleservernodeinH. V. NUMERICAL RESULTS In this section we evaluate the proposed algorithm using sample topologies following the Power Law. These topologies are taken from theBRITEpackage=-=[30]-=-byusingtheRouter Barabasi-Albert Model. We used random parameters for the network, rather than specific application parameters in order to test the overall performance of the algorithm under general c... |

100 | Greedy facility location algorithms analyzed using dual fitting with factorrevealing LP
- Jain, Mahdian, et al.
- 2003
(Show Context)
Citation Context ... ln |V | compared to the optimal. In our case we have arbitrary integer node loads gi and the functions fj(y) are subadditive and can be computed exactly only in pseudopolynomial time. As observed in =-=[22]-=- the assumption of unit loads can be removed by considering a modified network where node i is replaced with gi nodes each having the same links (and link costs). However, now the algorithm becomes ps... |

95 | Constrained mirror placement on the Internet
- Jamin, Jin, et al.
- 2001
(Show Context)
Citation Context ...ting clients’ latency constraints and servers’ capacity constraints by self organizing these replicas into a dissemination tree with small delay and bandwidth consumption for update dissemination. In =-=[8]-=- the authors considered the problem of placing a set of mirrors only at certain locations such that the maximum distance from a client to its closest mirror (server replica), based on round trip time,... |

78 | Improved approximation algorithms for capacitated facility location problems
- Chudak, Williamson
- 2005
(Show Context)
Citation Context ...olynomial time approximation algorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij (=-=[12]-=-, [13], [14], [15], [16], [17], [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality d... |

72 | Selection algorithms for replicated web servers
- Sayal, Breitbart, et al.
- 1998
(Show Context)
Citation Context ... cost for placing a mirror and showed that placing mirrors beyond a certain number offered little performance gain. Sayal et al presented some selection algorithms to access replicated Web servers in =-=[8]-=-. The algorithms found the closest replicated server for a client based on different metrics such as hop counts, round trip time and the HTTP request latency. In [9] the objective was to minimize the ... |

65 | Dynamic replica placement for scalable content delivery
- Chen, Katz, et al.
- 2013
(Show Context)
Citation Context ...as defined as the delay, hop count or the economic cost of the path between two nodes. They compared several heuristic solutions and found that a greedy algorithm had the best performance. Chen et al =-=[6]-=- tackled the replica placement problem from an other angle: minimizing the number of replicas when meeting clients’ latency constraints and servers’ capacity constraints by using a dissemination tree.... |

63 |
On the optimal placement of web proxies
- Li, Golin, et al.
- 1999
(Show Context)
Citation Context ...ast in several papers. Krishnan et al [1] developed polynomial optimal solutions to place a given number of servers in a tree network to minimize the average retrieval cost 1sof all clients. Li et al =-=[2]-=- investigated the placement of a limited number of Web proxies in a tree so that the overall latency for accessing the Web server is minimized. In [3] two objectives were studied: minimization of the ... |

58 | A simple efficient approximation scheme for the restricted shortest path problem
- Lorenz, Raz
- 1999
(Show Context)
Citation Context ...me approximation algorithm for the problem of finding the minimum constrained path from a source to a given destination was developed by Hassin [23]. An improvement of this algorithm was presented in =-=[24]-=-. The latter algorithm can be used as a basis for the development of a fully polynomial time approximation algorithm for Problem 1. We skip the details here and mention that the adaptation consists in... |

41 | Facility location with nonuniform hard capacities
- Pál, Tardos, et al.
- 2001
(Show Context)
Citation Context ...tion algorithms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], [13], [14], [15], =-=[16]-=-, [17], [18]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold in our case... |

38 | Hop-by-hop quality of service routing
- Mieghem, Neve, et al.
- 2001
(Show Context)
Citation Context ...known. There are fully polynomial time algorithms for computing these quantities, however in this section we will concentrate on efficient pseudopolynomial algorithms that work well in practice [19], =-=[20]-=-. We provide the discussion for Fij(d), since the same holds for Bij(d). The algorithm in [19], [20] are based on the fact that Fij(d) is a right continuous non-increasing step function with a finite ... |

28 | On Replica Placement for QoSAware Content Distribution
- Tang, Xu
(Show Context)
Citation Context ...o access replicated Web servers in [8]. The algorithms found the closest replicated server for a client based on different metrics such as hop counts, round trip time and the HTTP request latency. In =-=[9]-=- the objective was to minimize the amount of resources, storage and update, required to achieve a certain level of service. They assumed that all servers in the network are organized into a tree struc... |

24 | The Facility Location Problem with General Cost Functions
- Hajiaghayi, Mahdian, et al.
(Show Context)
Citation Context ...hms can be developed. Regarding Problem 3, there is an extensive work in the 6 i∈Vsliterature under various assumptions on the function fj(y) and on the costs cij ([12], [13], [14], [15], [16], [17], =-=[18]-=-). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold in our case. In the nex... |

9 | Efficient QoS Routing - Siachalou, Georgiadis - 2003 |

6 | A sparse knapsack algo-tech-cuit and its synthesis
- Andonov, Rajopadhye
- 1994
(Show Context)
Citation Context ...on. The function f(y) has properties similar to those of Fij(d) andBij(d). Specifically, it is a right continuous non-decreasing step function. Moreover, based on (7) and using an approach similar to =-=[21]-=-, an efficient pseudopolynomial algorithm can be developed for finding the jump points of f(y). Again, an important property in our case is that in one run of the algorithm, all jump points of f(y) up... |

5 |
Optimal placement of web proxies for replicated web servers in the internet
- Jia, Li, et al.
- 2001
(Show Context)
Citation Context ...average retrieval cost 1sof all clients. Li et al [2] investigated the placement of a limited number of Web proxies in a tree so that the overall latency for accessing the Web server is minimized. In =-=[3]-=- two objectives were studied: minimization of the overall access cost by all clients to access the Web site and minimization of the longest delay for any client to access the Web site. The problem was... |

5 | Placement of web-server proxies with consideration of read and update operations on the internet
- Jia, Li, et al.
- 2003
(Show Context)
Citation Context ... and minimization of the longest delay for any client to access the Web site. The problem was reduced to the placement of proxies in a set of trees whose root nodes are the server replicas. Jia et al =-=[4]-=- took the read and update operations into consideration. Qiu et al [5] also assumed a restricted number of replicas and no restriction of the number of requests served by each replica. A client could ... |

4 | Efficient QoS Routing - Siachalou, Georgiadis - 2003 |

4 |
The minimal cost distribution tree problem of recursive expiration-based consistency management
- Tang, Chanson
- 2004
(Show Context)
Citation Context ...d at the origin server. The construction of a distribution tree for a given set of replicas with the objective of minimizing the total communication cost of consistency management has been studied in =-=[11]-=-. Tang et. al. [12] presented a theoretical study on geographical replication of dynamic Web contents with the objective of minimizing the consistency management costs in terms of update transfers and... |

1 |
Constrained mirror palcement on the internet
- Jamin, Jiu, et al.
- 2001
(Show Context)
Citation Context ...kled the replica placement problem from an other angle: minimizing the number of replicas when meeting clients’ latency constraints and servers’ capacity constraints by using a dissemination tree. In =-=[7]-=- the authors considered the problem of placing a set of mirrors only at certain locations such that the maximum distance from a client to its closest mirror (server replica), based on round trip time,... |

1 |
A dynamic scalable and efficient content distribution network
- “SCAN
- 2002
(Show Context)
Citation Context ...ined as the delay, hop count or the economic cost of the path between two nodes. They compared several heuristic solutions and found that a greedy algorithm had the best performance. Chen et. al. [6]-=-=[7]-=- tackled the replica placement problem from another angle: minimizing the number of replicas while meeting clients’ latency constraints and servers’ capacity constraints by self organizing these repli... |

1 |
cost replication of dynamic web contents under flat update delivery
- “Minimal
- 2004
(Show Context)
Citation Context ...ver. The construction of a distribution tree for a given set of replicas with the objective of minimizing the total communication cost of consistency management has been studied in [11]. Tang et. al. =-=[12]-=- presented a theoretical study on geographical replication of dynamic Web contents with the objective of minimizing the consistency management costs in terms of update transfers and 2sobject reconstru... |

1 |
Approximation algorithms for the metricfacilitylocationandk-median problems using the primal-dual schema and the lagrangian relaxation
- Jain, Vazirani
- 2001
(Show Context)
Citation Context ...me approximation algorithms can be developed. Regarding Problem 3, there is an extensive work in the literature under various assumptions on the function fj(y) and on the costs cij ([16], [17], [18], =-=[19]-=-, [20], [21], [22]). Most of the work is concentrated on the case of “metric” costs, i.e., it is assumed that costs satisfy the inequality cij + cjk ≥ cik. However, this inequality does not hold in ou... |