Dynamic replication to improve input/output scalability of genomic alignment
Palghat P. Vaidyanathan, Tara M. Madhyastha · 2004
Many scientific applications are I/O-intensive, whichmakes optimization and scaling difficult, especially on parallelarchitectures. The I/O requirements of computationalbiology applications are different from other scientific applications;many are embarrassingly parallel and requirerepeated read-only access to a large global database. Inthe first part of this paper we examine the scalability ofan embarrassingly parallel computational biology application:psLayout, which played a crucial role in the mappingof the human genome. Input file location is a importantfactor affecting the scalability and performance of this application.We designed a library for automatic data replicationto address this issue. In the second part of this paperwe describe our user-level library for location-transparentstorage.