1 Optimal Online Algorithms for Minimax Resource Scheduling Imen BOURGUIBA CAS 744 McMaster University.

Slides:



Advertisements
Similar presentations
Optimal Algorithms for k-Search with Application in Option Pricing Julian Lorenz, Konstantinos Panagiotou, Angelika Steger Institute of Theoretical.
Advertisements

Multicriteria Decision-Making Models
Turing Machines January 2003 Part 2:. 2 TM Recap We have seen how an abstract TM can be built to implement any computable algorithm TM has components:
Measuring Time Complexity
Instantly Decodable Network Codes for Real-Time Applications Anh Le, Arash Tehrani, Alex Dimakis, Athina Markopoulou UC Irvine, USC, UT Austin Presented.
Class-constrained Resource Allocation Problems Tami Tamir Thesis advisor: Hadas Shachnai.
Machine scheduling Job 1Job 3 Job 4 Job 5Machine 1 Machine 2 time 0C max Job 2.
1 1 Slide Chapter 1 & Lecture Slide Body of Knowledge n Management science Is an approach to decision making based on the scientific method Is.
Online Algorithm Huaping Wang Apr.21
Tight Bounds for Online Class- constrained Packing Hadas Shachnai Bell Labs and The Technion IIT Tami Tamir The Technion IIT.
Algorithm Design Methods Spring 2007 CSE, POSTECH.
Load Balancing Parallel Applications on Heterogeneous Platforms.
Routing and Congestion Problems in General Networks Presented by Jun Zou CAS 744.
Problems and Their Classes
Chapter 11 Limitations of Algorithm Power Copyright © 2007 Pearson Addison-Wesley. All rights reserved.
Scheduling Planning with Actions that Require Resources.
13-Optimization Assoc.Prof.Dr. Ahmet Zafer Şenalp Mechanical Engineering Department Gebze Technical.
S YSTEM -W IDE E NERGY M ANAGEMENT FOR R EAL -T IME T ASKS : L OWER B OUND AND A PPROXIMATION Xiliang Zhong and Cheng-Zhong Xu ICCAD 2006, ACM Trans. on.
CPE555A: Real-Time Embedded Systems
Properties of SPT schedules Eric Angel, Evripidis Bampis, Fanny Pascual LaMI, university of Evry, France MISTA 2005.
1 Modeling and Optimization of VLSI Interconnect Lecture 9: Multi-net optimization Avinoam Kolodny Konstantin Moiseev.
Online Scheduling with Known Arrival Times Nicholas G Hall (Ohio State University) Marc E Posner (Ohio State University) Chris N Potts (University of Southampton)
Train DEPOT PROBLEM USING PERMUTATION GRAPHS
Complexity 7-1 Complexity Andrei Bulatov Complexity of Problems.
Complexity 15-1 Complexity Andrei Bulatov Hierarchy Theorem.
The Cache Location Problem IEEE/ACM Transactions on Networking, Vol. 8, No. 5, October 2000 P. Krishnan, Danny Raz, Member, IEEE, and Yuval Shavitt, Member,
Adaptive Data Collection Strategies for Lifetime-Constrained Wireless Sensor Networks Xueyan Tang Jianliang Xu Sch. of Comput. Eng., Nanyang Technol. Univ.,
Complexity 5-1 Complexity Andrei Bulatov Complexity of Problems.
1 Single Machine Deterministic Models Jobs: J 1, J 2,..., J n Assumptions: The machine is always available throughout the scheduling period. The machine.
Math443/543 Mathematical Modeling and Optimization
1 Ecole Polytechnque, Nov 7, 2007 Scheduling Unit Jobs to Maximize Throughput Jobs:  all have processing time (length) = 1  release time r j  deadline.
Online Algorithms Motivation and Definitions Paging Problem Competitive Analysis Online Load Balancing.
Energy-Efficient Rate Scheduling in Wireless Links A Geometric Approach Yashar Ganjali High Performance Networking Group Stanford University
Dynamic lot sizing and tool management in automated manufacturing systems M. Selim Aktürk, Siraceddin Önen presented by Zümbül Bulut.
The k-server Problem Study Group: Randomized Algorithm Presented by Ray Lam August 16, 2003.
1 Worst-Case Equilibria Elias Koutsoupias and Christos Papadimitriou Proceedings of the 16th Annual Symposium on Theoretical Aspects of Computer Science.
Computability and Complexity 24-1 Computability and Complexity Andrei Bulatov Approximation.
Recap Priorities task-level static job-level static dynamic Migration task-level fixed job-level fixed migratory Baker/ Oh (RTS98) Pfair scheduling This.
Evaluation of Algorithms for the List Update Problem Suporn Pongnumkul R. Ravi Kedar Dhamdhere.
Linear Programming Econ Outline  Review the basic concepts of Linear Programming  Illustrate some problems which can be solved by linear programming.
Linear programming. Linear programming… …is a quantitative management tool to obtain optimal solutions to problems that involve restrictions and limitations.
Introduction to Quantitative Business Methods (Do I REALLY Have to Know This Stuff?)
An equivalent version of the Caccetta-Häggkvist conjecture in an online load balancing problem Angelo Monti 1, Paolo Penna 2, Riccardo Silvestri 1 1 Università.
Resource Allocation for E-healthcare Applications
Yossi Azar Tel Aviv University Joint work with Ilan Cohen Serving in the Dark 1.
Linear Programming Topics General optimization model LP model and assumptions Manufacturing example Characteristics of solutions Sensitivity analysis Excel.
Online Paging Algorithm By: Puneet C. Jain Bhaskar C. Chawda Yashu Gupta Supervisor: Dr. Naveen Garg, Dr. Kavitha Telikepalli.
Introduction A GENERAL MODEL OF SYSTEM OPTIMIZATION.
Michael J. Neely, University of Southern California CISS, Princeton University, March 2012 Asynchronous Scheduling for.
§1.4 Algorithms and complexity For a given (optimization) problem, Questions: 1)how hard is the problem. 2)does there exist an efficient solution algorithm?
MULTICELL UPLINK SPECTRAL EFFICIENCY OF CODED DS- CDMA WITH RANDOM SIGNATURES By: Benjamin M. Zaidel, Shlomo Shamai, Sergio Verdu Presented By: Ukash Nakarmi.
Linear Program Set Cover. Given a universe U of n elements, a collection of subsets of U, S = {S 1,…, S k }, and a cost function c: S → Q +. Find a minimum.
Approximation Algorithms Department of Mathematics and Computer Science Drexel University.
Frequency Capping in Online Advertising Moran Feldman Technion Joint work with: Niv Buchbinder,The Open University of Israel Arpita Ghosh,Yahoo! Research.
Rounding scheme if r * j  1 then r j := 1  When the number of processors assigned in the continuous solution is between 0 and 1 for each task, the speed.
Improved Competitive Ratios for Submodular Secretary Problems ? Moran Feldman Roy SchwartzJoseph (Seffi) Naor Technion – Israel Institute of Technology.
© The McGraw-Hill Companies, Inc., Chapter 12 On-Line Algorithms.
Non-Preemptive Buffer Management for Latency Sensitive Packets Moran Feldman Technion Seffi Naor Technion.
Introduction to Quantitative Business Methods (Do I REALLY Have to Know This Stuff?)
11 -1 Chapter 12 On-Line Algorithms On-Line Algorithms On-line algorithms are used to solve on-line problems. The disk scheduling problem The requests.
On the behaviour of an edge number in a power-law random graph near a critical points E. V. Feklistova, Yu.
Moran Feldman The Open University of Israel
Linear Programming Topics General optimization model
Policy Gradient in Continuous Time
A new and improved algorithm for online bin packing
Constraint management
György Dósa – M. Grazia Speranza – Zsolt Tuza:
Graphical solution A Graphical Solution Procedure (LPs with 2 decision variables can be solved/viewed this way.) 1. Plot each constraint as an equation.
Dr. Arslan Ornek DETERMINISTIC OPTIMIZATION MODELS
Chapter 1. Formulations.
Presentation transcript:

1 Optimal Online Algorithms for Minimax Resource Scheduling Imen BOURGUIBA CAS 744 McMaster University

2 Outline Introduction Objectives The ORMP problem The HLBP problem Comparison Problem definition Analysis Randomized algorithms Conclusion

3 Introduction The objective is to minimize the maximum level of resource allocated at any time during the planning period, this problem is called Online Resource Minimization Problem (ORMP) The objective is to minimize the maximum amount of work assigned to any machine, the problem is called the Hierarchical Line Balancing Problem (HLBP)

4 Introduction Both ORMP and HLBP are special cases of a more general problem, the Online Min- Max Problem (OMMP) The quality of the algorithm is evaluated by the competitive ratio

5 Objectives A simple parameterized deterministic algorithm, called the  -policy, with parameter  and competitive ratio , provided it produces a feasible solution The  -policy is also optimal among all randomized algorithms

6 The ORMP problem Work with different deadlines arrives over time and has to be performed using a resource. The quantities of work that arrive as well as their deadlines become known only at the times of arrival. At a given set of time points, the decision maker decides how much resource to allocate and what part of the available work to perform at that time

7 The HLBP problem Work with different requirements arrives over time and has to be assigned to a collection of machines with different capabilities. The machines form a linear hierarchy based on their capabilities The amount of work that arrives as well as the required machine capabilities become known only at the time of arrival

8 Comparison ORMP appears to be a new problem, and there is no existing literature discussing it For HLBP, the optimal competitive ratio tends to e when the number of machines goes to infinity. When the machines have same capability, the competitive ratio is 2/1-m for the m identical parallel machines [Graham]

9 Problem definition An instance  of the OMMP is a finite sequence (a(1), ….a(T)) of length T The set of all instances with parameter  is denoted by   For example, the set of all instances of length T is denoted by  T let  t = (a(1 ), ….a(T)) denote the first t elements of instance  ; that is,  t denotes the history of instance  up to time t<=T For any instance of length T, a solution r is a sequence (r(1), …...r(T))  R + T of T nonnegative real numbers

10 Problem definition  (  )(t): the decision at time t for instance  under  r  (t) denotes the decision  (  )(t) when  is fixed For any instance    T and any deterministic algorithm  the value   (  ) = max {r  (1), r  (2), …, r  (t)} and the optimal value is  * (  ) = inf r ( {max {r(1), r(2), …, r(t)}} The  -policy is an online policy with parameter  >= 1, defined by: r i  (  ) =   * (  i ) for all i    is the worst-case competitive ratio of policy  over all instances with time     = Sup  {   (  )/  * (  )} An instance  of the OMMP is a finite sequence (a(1), ….a(T)) of length T The set of all instances with parameter  is denoted by   For example, the set of all instances of length T is denoted by  T let  t = (a(1 ), ….a(T)) denote the first t elements of instance  ; that is,  t denotes the history of instance  up to time t For any instance of length T, a solution r is a sequence (r(1), …...r(T))  R + T of T nonnegative real numbers

11 Analysis A simple parameterized algorithm, called the  -policy, with parameter   and competitive ratio  , provided that it is feasible To be feasible a solution must satisfy 3 constraints: 1) the total amount of work performed at time i cannot exceed the amount of work accomplished with r i resources 2) all work must be performed with the respective deadlines 3) the work cannot be performed before it has arrived

12 Theorem For any algorithm    DO if    <  then the  -policy   with parameter   =    achieves the same competitive ratio,    =    Proof:

13 Theorem (cont)  (  )(t): the decision at time t for instance  under  r  (t) denotes the decision  (  )(t) when  is fixed For any instance    T and any deterministic algorithm  the value   (  ) = max {r  (1), r  (2), …, r  (t)} and the optimal value is  * (  ) = inf r ( max {r(1), r(2), …, r(t)}} The  -policy is an online policy with parameter  >= 1, defined by: r i  (  ) =   * (  i ) for all i

14 Randomized algorithms For some problems, randomized algorithms can have better competitive ratios than deterministic algorithms (Motwani and Raghavan and Hoogeveen and Vestjens) A randomized algorithm can have better competitive ratios for the OMMP than any deterministic algorithm.

15 Conclusion The  -policy theory developed in this paper is a powerful tool for finding worst-case optimal algorithms for online min-max problems With an appropriate choice of parameters, the  -policy has as good a competitive ratio as any other deterministic algorithm Under mild conditions an optimal parameter value exists, so the  -policy is

16 References B. Hunsaker y, A. J. Kleywegt, M. W. P. Savelsbergh, andC. A. Tovey, Optimal Online Algorithms for Minimax Resource Scheduling, SIAM J. Discrete Math. Vol. 16, No. 4, 2003, pp A. J. Kleywegt, V. S. Nori, M. W. P. Savelsbergh, and C. A. Tovey, Online resource minimization, in Proceedings of the Tenth Annual ACM-SIAM Symposium on Discrete Algorithms, Baltimore, MD, January 1999, SIAM, Philadelphia, 1999, pp