## Competitive Algorithms for Distributed Data Management (0)

### Cached

### Download Links

- [www.icsi.berkeley.edu]
- [condor.cs.jhu.edu]
- [www.cs.technion.ac.il]
- [www.cs.technion.ac.il]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proceedings of the 24th Annual ACM Symposium on Theory of Computing |

Citations: | 100 - 8 self |

### BibTeX

@INPROCEEDINGS{Bartal_competitivealgorithms,

author = {Yair Bartal and Amos Fiat and Yuval Rabani},

title = {Competitive Algorithms for Distributed Data Management},

booktitle = {In Proceedings of the 24th Annual ACM Symposium on Theory of Computing},

year = {},

pages = {39--50}

}

### Years of Citing Articles

### OpenURL

### Abstract

We deal with the competitive analysis of algorithms for managing data in a distributed environment. We deal with the file allocation problem ([DF], [ML]), where copies of a file may be be stored in the local storage of some subset of processors. Copies may be replicated and discarded over time so as to optimize communication costs, but multiple copies must be kept consistent and at least one copy must be stored somewhere in the network at all times. We deal with competitive algorithms for minimizing communication costs, over arbitrary sequences of reads and writes, and arbitrary network topologies. We define the constrained file allocation problem to be the solution of many individual file allocation problems simultaneously, subject to the constraints of local memory size. We give competitive algorithms for this problem on the uniform network topology. We then introduce distributed competitive algorithms for on-line data tracking (a generalization of mobile user tracking [AP1...

### Citations

729 |
Tarjan. Amortized efficiency of list update and paging rules
- Sleator, E
- 1985
(Show Context)
Citation Context ... A potential function \Phi is called strict iff \Phi 0 = \Phi(h 0 ; A 0 ) = 0. Potential functions are useful in the competitive analysis of on-line algorithms, as shown in the following theorem (see =-=[ST]-=-): Theorem 1 If there exists a potential function for Alg (and c), then Alg is c-competitive (against adversaries for which property 2 above holds). The following types of potential functions are comm... |

207 | Online tracking of mobile users
- Awerbuch, Peleg
- 1995
(Show Context)
Citation Context ... We manage a distributed data structure that allows fast access to the closest copy of a file, while the cost of managing the data structure is amortized against the cost of the data movement itself. =-=[AP1]-=- solve a similar problem, they allow a move operator to be applied to a mobile user, but do not efficiently support birth and death. We allow insert and delete operations. The competitive ratio is pol... |

185 | An optimal on-line algorithm for metrical task system - Borodin, Linial, et al. - 1992 |

164 | Competitive paging algorithms - Fiat, Karp, et al. - 1991 |

157 |
Comparative Models of the File Assignment Problem
- Dowdy, Foster
- 1982
(Show Context)
Citation Context ... University Tel-Aviv 69978, Israel. Yuval Rabani Abstract We deal with the competitive analysis of algorithms for managing data in a distributed environment. We deal with the file allocation problem (=-=[DF]-=-, [ML]), where copies of a file may be be stored in the local storage of some subset of processors. Copies may be replicated and discarded over time so as to optimize communication costs, but multiple... |

142 |
Dynamic Steiner tree problem
- Imase, Waxman
- 1991
(Show Context)
Citation Context ...tree problem is equivalent to a special case of the file allocation problem, where D = 1, only read requests are issued, and the algorithm is forced to replicate upon a read request. Imaze and Waxman =-=[IW]-=- have defined this problem and gave upper and lower bounds for it. They have shown that the greedy on-line Steiner tree algorithm is dlog ne competitive. 13 For completeness of the discussion we give ... |

140 |
Competitive algorithms for on-line problems
- Manasse, McGeoch, et al.
- 1988
(Show Context)
Citation Context ...t) + some additive constant. A competitive algorithm with a competitive ratio of c is called strictly competitive if the additive constant is zero. Models for on-line problems are presented in [BLS], =-=[MMS]-=-, [BBKTW]. Competitive analysis of distributed data management algorithms begins with Karlin et. al. in [KMRS] who analyze competitive algorithms for snoopy caching on a bus connected PRAM. If the on-... |

138 | On the power of randomization in on-line algorithms
- Ben-David, Borodin, et al.
- 1994
(Show Context)
Citation Context ...me additive constant. A competitive algorithm with a competitive ratio of c is called strictly competitive if the additive constant is zero. Models for on-line problems are presented in [BLS], [MMS], =-=[BBKTW]-=-. Competitive analysis of distributed data management algorithms begins with Karlin et. al. in [KMRS] who analyze competitive algorithms for snoopy caching on a bus connected PRAM. If the on-line algo... |

136 |
Sparse partitions
- Awerbuch, Peleg
- 1990
(Show Context)
Citation Context ...e solution to the data tracking problem (Section 9.2.1), we make use two tools. One is a graph-theoretic structure of regional matchings, given by [AP1], an application of the sparse graph partitions =-=[AP2]-=-. An m-regional matching is an assignment of 2 sets of processors to each processor, a read-set and a write-set, such that for every two processors p and q that satisfy d(p; q)sm, the read-set of p an... |

86 |
Memory versus randomization in on-line algorithms
- Raghavan, Snir
- 1994
(Show Context)
Citation Context ...e also prove that \Omega\Gammaat/ n) is the best competitive ratio one can obtain for general networks, even for randomized algorithms against an oblivious adversary. Our algorithm is also memoryless =-=[RS]-=- (I.e., its decisions depend only on its current configuration and the current request). We give an optimal deterministic 3competitive algorithm for the uniform architecture (e.g., bus based). Westbro... |

73 | New results on server problems
- Chrobak, Karloff, et al.
- 1991
(Show Context)
Citation Context ...l Deterministic Algorithm for Trees In appendix C we describe a nearly optimal algorithm for processors in a tree topology. The algorithm is simple and intuitive, and is inspired by the algorithms of =-=[CKPV]-=-, and [CL]. The algorithm is presented in two stages. First we give an algorithm for the continuous case (CT), i.e. when every point along an edge may hold a copy of the file. Then we show how the ide... |

63 | Optimal File Allocation in a Multiple Computer System - Chu - 1969 |

62 |
Competitive algorithms for replication and migration poblems
- Black, Sleator
- 1989
(Show Context)
Citation Context ...ge. We consider the competitive performance [ST, KMRS, MMS, BLS, BBKTW] of algorithms for these problems, and present algorithms with an optimal or nearly optimal competitive ratio. Black and Sleator =-=[BS]-=- consider competitive algorithms for two partial components of the file allocation family of problems. Our file allocation problem may be viewed as the combined solution to the two subproblems defined... |

50 | On-line Steiner trees in the Euclidean plane
- Alon, Azar
- 1992
(Show Context)
Citation Context ...is dlog ne competitive. 13 For completeness of the discussion we give here an alternative very simple proof of this claim. Analysis of the greedy Steiner tree algorithm was also independently made by =-=[AA]-=-, [ABF1], [CV] and [WY2] giving similar bounds. The Greedy Steiner Tree Algorithm. The greedy Steiner tree algorithm connects a new point to the closest point already in the tree. Theorem 4 The greedy... |

46 | The server problem and on-line games
- Chrobak, Larmore
- 1992
(Show Context)
Citation Context ...algorithms for simpler subproblems. Our analysis of the competitive file allocation algorithm is based upon the natural potential function for on-line Steiner tree algorithms. Similar ideas appear in =-=[CL]-=- in the context of deterministic algorithms, our natural potential function is defined for both deterministic and randomized algorithms. 1.3 The Constrained File Allocation Problem If it is not true t... |

29 | Randomized algorithms for multiprocessor page migration
- Westbrook
- 1992
(Show Context)
Citation Context ...e processor that holds the copy. [BS] call this problem the file migrationsproblem. [BS] give an optimal 3-competitive ratio for this problem on the uniform network topology and for trees. Westbrook (=-=[W]-=-) gives a randomized 3-competitive algorithm against an adaptive on-line adversary for any network, and a 1+ OE-competitive randomized algorithm against an oblivious adversary. The data migration prob... |

22 | Distance-dependent distributed directories - Peleg - 1993 |

18 | Heat & dump: Competitive distributed paging
- Awerbuch, Bartal, et al.
- 1993
(Show Context)
Citation Context ... Theorem SBA is still open. Motivated by the famous [MMS] conjecture, we conjecture that the constrained file allocation problem has a deterministic competitive ratio of O(m) on arbitrary topologies. =-=[ABF2]-=- gives an O(log m)-competitive randomized algorithm for the constrained file allocation problem on the uniform network. We hazard the guess that similar results can be obtained by randomized algorithm... |

9 | Optimal Multiprocessor Migration Algorithms Using Work Functions - Chrobak, Larmore, et al. - 1989 |

9 |
Constructing reliable communication networks of small wight online. unpublished manuscript
- Chandra, Vishwanathan
- 1993
(Show Context)
Citation Context ...petitive. 13 For completeness of the discussion we give here an alternative very simple proof of this claim. Analysis of the greedy Steiner tree algorithm was also independently made by [AA], [ABF1], =-=[CV]-=- and [WY2] giving similar bounds. The Greedy Steiner Tree Algorithm. The greedy Steiner tree algorithm connects a new point to the closest point already in the tree. Theorem 4 The greedy Steiner tree ... |

7 |
An Optimal On-Line k-Server Algorithm for Trees
- Chrobak, Larmore
- 1991
(Show Context)
Citation Context ...tic Algorithm for Trees In appendix C we describe a nearly optimal algorithm for processors in a tree topology. The algorithm is simple and intuitive, and is inspired by the algorithms of [CKPV], and =-=[CL]-=-. The algorithm is presented in two stages. First we give an algorithm for the continuous case (CT), i.e. when every point along an edge may hold a copy of the file. Then we show how the ideas from th... |

5 |
Greedy On-Line Steiner Tree and Generalized Steiner Problems
- Yan
- 1993
(Show Context)
Citation Context ... 13 For completeness of the discussion we give here an alternative very simple proof of this claim. Analysis of the greedy Steiner tree algorithm was also independently made by [AA], [ABF1], [CV] and =-=[WY2]-=- giving similar bounds. The Greedy Steiner Tree Algorithm. The greedy Steiner tree algorithm connects a new point to the closest point already in the tree. Theorem 4 The greedy Steiner tree algorithm ... |

5 | Locality-sensitive resource allocation - Awerbuch, Peleg - 1990 |

1 |
Competitive Distributed File Allocation. To Appear
- Awerbuch, Bartal, et al.
- 1993
(Show Context)
Citation Context ...g ne competitive. 13 For completeness of the discussion we give here an alternative very simple proof of this claim. Analysis of the greedy Steiner tree algorithm was also independently made by [AA], =-=[ABF1]-=-, [CV] and [WY2] giving similar bounds. The Greedy Steiner Tree Algorithm. The greedy Steiner tree algorithm connects a new point to the closest point already in the tree. Theorem 4 The greedy Steiner... |

1 | Computer Architecture: A Quantitative Aproach - Hennessy, Patterson - 1990 |