Bin Packing Problem with Rejection in the Information Organization and Allocation of Internet and Web

何勇,谈之奕,任峰

DOI: https://doi.org/10.3321/j.issn:0254-4164.2003.12.023

2003-01-01

Chinese Journal of Computers

Abstract:In this paper, we consider the following bin packing problem with rejection. We are given n items, each with a capacity and penalty, and unlimited bins with equal capacity. An item can be either rejected, in which case we pay its penalty, or put in one of the bins, in which case it causes the decreasing of the capacity of a certain bin. The objective is to minimize the sum of the number of the used bins and the penalties of all rejected items. This is a new combinatorial optimization problem which has many applications in practice, such as in organizing and allocating information on web servers. We first provide a lower bound of its optimal value which can be used to solve the problem optimally by branch and bound method. Since the problem is strongly NP-hard, we further consider the design and analysis of its online and offline approximation algorithms. We present an offline algorithm with absolute worst case ratio 2, and an online algorithm with asymptotic worst case ratio 2 and absolute worst case ratio no greater than 3. Moreover we provide lower worst-case bounds of any online and offline algorithms.