Large Scale Network-Centric Distributed Systems

20

Broker-Mediated Cloud-Aggregation Mechanism Using Markovian Queues for Scheduling Bag-of-Tasks (BoT) Applications

Ganesh Neelakanta Iyer and Bharadwaj Veeravalli

Contents

20.1 Introduction

20.2 Literature Review and Contributions

20.2.1 Literature Review

20.2.2 Contributions and Scope of This Chapter

20.3 Problem Setting and Notations

20.4 Proposed Cloud Aggregation Mechanism

20.4.1 Task Distribution to Minimize the Application Completion Time

20.4.1.1 Integer Approximation Techniques

20.4.1.2 Eliminating CSPs with Lower Resource Capabilities

20.4.2 Task Distribution Based on Budget Requirements

20.5 Performance Evaluation and Discussions

20.5.1 Analysis of Task Execution Time vs. Budget Requirements

20.5.2 Analysis of the Total User Expenditure vs. Budget Requirements

20.5.3 Analysis of Task Distribution Based on Budget Requirements

20.6 Discussions

20.6.1 Applicability of Our Model to Divisible Load Applications

20.6.2 Flexibility in Considering More User Requirements

20.6.3 Consideration of Cloud Characteristics

20.7 Conclusion

Reference

20.1 Introduction

Many data-intensive applications from several fields, including data mining [1] and computational biology [2], have been modeled as Bag-of-Tasks (BoT) applications. BoT applications are parallel applications whose tasks are independent of each other [3], [4]. In other words, a BoT application is an application that executes in terms of tasks [5]. Examples include simulation of a large number of independent trials or events ...