| Popis: |
Commercial cloud systems allow massively parallel execution of a computing task for little money. We want to exploit this economic opportunity by solving classical problems in Operations Research through complete enumeration, especially if these problems can be expressed as integer programming problems. We propose and evaluate here a data structure, Scalable Virtual Distributed Hashing, that autonomously extends the computing task over as many nodes as are needed in order return a result within a time limit set by the user. Our data structure deals with varying and changing node capacities and the effects of node failures. It is modeled after Scalable Distributed Data Structures and Extendible Hashing in particular. |
