The grid vision of a single computing utility has yet to materialize: while many grids with thousands of processors each exist, most work in isolation. An important obstacle for the effective and efficient inter-operation of grids is the problem of resource selection. In this paper we propose a solution to this problem that combines the hierarchical and decentralized approaches for interconnecting grids. In our solution, a hierarchy of grid sites is augmented with peer-to-peer connections between sites under the same administrative control. To operate this architecture, we employ the key concept of delegated matchmaking, which temporarily binds resources from remote sites to the local environment. With trace-based simulations we evaluate our solution under various infrastructural and load conditions, and we show that it outperforms other approaches to inter-operating grids. Specifically, we show that delegated matchmaking achieves up to 60 % more goodput and completes 26 % more jobs than its best alternative, daily. 1

We introduce intelligent web caching algorithms that employ predictive models of web requests; the general idea is to extend the LRU policy of web and proxy servers by making it sensible to web access models extracted from web log data using data mining techniques. Two approaches have been studied in particular, frequent patterns and decision trees. The experimental results of the new algorithms show substantial improvement over existing LRU-based caching techniques, in terms of hit rate. We designed and developed a prototypical system, which supports data warehousing of web log data, extraction of data mining models and simulation of the web caching algorithms.

Abstract—Power consumption imposes a significant cost for data centers implementing cloud services, yet much of that power is used to maintain excess service capacity during periods of predictably low load. This paper investigates how much can be saved by dynamically ‘right-sizing ’ the data center by turning off servers during such periods, and how to achieve that saving via an online algorithm. We prove that the optimal offline algorithm for dynamic right-sizing has a simple structure when viewed in reverse time, and this structure is exploited to develop a new ‘lazy ’ online algorithm, which is proven to be 3-competitive. We validate the algorithm using traces from two real data center workloads and show that significant cost-savings are possible. I.

In this paper, we study online algorithms when the input is not chosen adversarially, but consists of draws from some given probability distribution. While this model has been studied for online problems like paging and k-server, it is not known how to beat the Θ(log n) bound for online Steiner tree if at each time instant, the demand vertex is a uniformly random vertex from the graph. For the online Steiner tree problem, we show that if each demand vertex is an independent draw from some probability distribution π: V → [0, 1], a variant of the natural greedy algorithm achieves Eω[A(ω)]/Eω[OPT(ω)] = O(1); moreover, this result can be extended to some other subadditive problems. Both assumptions that the input sequence consists of independent draws from π, and that π is known to the algorithm are both essential; we show (almost) logarithmic lower bounds if either assumption is violated. Moreover, we give preliminary results on extending the Steiner tree results above to the related “expected ratio” measure Eω[A(ω)/OPT(ω)]. Finally, we use these ideas to give an average-case analysis of the Universal TSP problem. 1

Motivated by an application in thinwire visualization, we study an abstract on-line scheduling problem where the size of each requested service can be scaled down by the scheduler. Thus our problem embodies a notion of "Level of Service" that is increasingly important in multimedia applications. We give two schedulers FirstFit and EndFit based on two simple heuristics, and generalize them into a class of greedy schedulers. We show that both FirstFit and EndFit are 2-competitive, and any greedy scheduler is 3-competitive. These bounds are shown to be tight.

Traditionally, the development of computing systems has been focused on performance improvements driven by the demand of applications from consumer, scientific and business domains. However, the ever increasing energy consumption of computing systems has started to limit further performance growth due to overwhelming electricity bills and carbon dioxide footprints. Therefore, the goal of the computer system design has been shifted to power and energy efficiency. To identify open challenges in the area and facilitate future advancements it is essential to synthesize and classify the research on power and energy-efficient design conducted to date. In this work we discuss causes and problems of high power / energy consumption, and present a taxonomy of energy-efficient design of computing systems covering the hardware, operating system, virtualization and data center levels. We survey various key works in the area and map them to our taxonomy to guide future design and development efforts. This chapter is concluded with a discussion of advancements identified in energy-efficient computing and our vision on future

Energy expenditure has become a significant fraction of data center operating costs. Recently, “geographical load balancing” has been suggested to reduce energy cost by exploiting the electricity price differences across regions. However, this reduction of cost can paradoxically increase total energy use. This paper explores whether the geographical diversity of Internet-scale systems can additionally be used to provide environmental gains. Specifically, we explore whether geographical load balancing can encourage use of“green”renewable energy and reduce use of “brown ” fossil fuel energy. We make two contributions. First, we derive two distributed algorithms for achieving optimal geographical load balancing. Second, we show that if electricity is dynamically priced in proportion to the instantaneous fraction of the total energy that is brown, then geographical load balancing provides significant reductions in brown energy use. However, the benefits depend strongly on the degree to which systems accept dynamic energy pricing and the form of pricing used.

The most common approach for modeling and solving routing and scheduling problems in a dynamic setting is to solve, as close to optimal as possible, a series of deterministic, myopic models. The argument is most often made that, if the data changes, then we should simply reoptimize. We use the setting of the load matching problem that arises in truckload trucking to compare the value of optimal myopic solutions versus varying degrees of greedy, suboptimal myopic solutions in the presence of three forms of uncertainty: customer demands, travel times, and, of particular interest, user noncompliance. A simulation environment is used to test different dispatching strategies under varying levels of system dynamism. An important issue we consider is that of user noncompliance, which is the effect of optimizing when users do not adopt all of the recommendations of the model. Our results show that (myopic) optimal solutions only slightly outperform greedy solutions under relatively high levels of uncertainty, and that a particular suboptimal solution actually outperforms optimal solutions under a wide range of conditions. The challenge of optimizing routing and scheduling problems in a real-time setting has been receiving

Discrete Optimization techniques have become a major and successful tool for modelling and solving many real world problems. In modelling real-time applications, we often have to face the inherent difficulty that an essential part of the data arrives sequentially in real-time, and that decision support is requested at the same time. Online and real-time algorithms are designed to handle such difficulties. We review some theoretical and practical aspects of online algorithms. Starting with theoretical concepts for performance evaluation, we survey typical results for classical optimization problems of discrete structure. Finally, we describe results on solving Discrete Optimization models for some real-time applications. Discrete Optimization has become a major and successful tool in modelling and solving real world problems arising in computer science, in economics, and in engineering. This success is based on two essential facts: computation times have decreased dramatically by the im...

When parallel join strategies are considered, good workload balancing methods are important in order to achieve reasonable performances. We study here parallel join algorithms that comprise some kind of load balancing activity. By taking into account different skew handling techniques, we discuss pros and cons of well known existing strategies and define a new taxonomy for this class of algorithms, according to the moment the load balancing technique is applied. The choice of a specific step of the algorithm to balance the load is closely related to the efficacy in handling the effects of different types of skew. Based on this idea, we introduce an alternate approach for parallel join processing, which tries to handle all skew situations by dynamically determining and assigning variablesized tasks to be executed at each processor node. Keywords: parallel databases, parallel join, workload balancing, variable-sized tasks, taxonomy. 1 Introduction The use of parallel algorithms to eva...