walsaip.uprm

1. 5 4 1 2 7 6 3 Tools Problem Formulation Proposed Solution References Conclusions Methodology Preliminary Results Gridjobs – A high throughput computing framework www.walsaip.uprm.edu John Sanabria – PhD Student Prof. Wilson Rivera – Advisor PDC Group, ECE Department, University of Puerto Rico, Email: john.sanabria@ece.uprm.eduMayagüez Campus How to orchestrate multiple services in grid environments to provide adaptivity under resource and service availability uncertainty? Gridjobs is a framework for developing and deploying different scheduling algorithms based on stochastic techniques. It was developed under Groovy/Grails which are programming environments that interact gracefully with Jar libraries and JDBC databases compliant. Client modules and other Gridjobs instances can interact through web service and RMI protocols. Data analysis is achieved using scripts developed in R. Grid System Model Uncertainty Resources are connected via two-level hierarchical networks. The first level is a wide area network that connects local area networks or virtual organizations at the second level. max E[f (x,y)] subject to: E[gj (x,y)] 0, j = 1, 2, . . . , p • The proposed solution intends to leverage resource utilization and provide efficient execution of service based applications. • Policy definitions at different levels allow: the users to direct the service execution and the resource to define its utilization mechanisms. • Similar proposals deal indistinctly with the provisioning and orchestration problem under controlled environments, whereas our model deals with uncertain factors, such as user demands and resource availability, among others. Gridjobs Data collected from continuous executions of a service used for signal processing over five different PRAGMA resources show high levels of variability in execution time. However, our aim is to adapt DLT models along with adequate forecasting techniques based on stochastic analysis to provide efficient and reliable execution plans for long running tasks. Gridjobs is a first level gateway composed of modules allowing the integration of different uncertain programming techniques to deal with unknown factors that affect resource availability in a grid environment. PRAGMA, our test-bed • Emerging platforms based on global infrastructures present new challenges to the reliable and efficient utilization of the resources due their uncertain availability. • Gridjobs offers a new platform for the execution of scientific applications providing a test-bed for the development of new scheduling techniques based on the performance profile of resources. D. Arias, J. Sanabria, W. Rivera “Grid Based Pervasive Distributed Storage”, IEEE International Symposium on Wireless Pervasive Computing (ISWPC), 2007 Rabin, M. O., “Efficient dispersal of information for security, load balancing, and fault tolerance”, Journal ACM. 36, 2 (Apr. 1989), 335-348. Plank, J. S., “A tutorial on Reed-Solomon coding for fault-tolerance in RAID-like systems”, Software-Practice and Experience (SPE), 27(9):995.1012, Sept. 1997. Correction in James S. Plank and Ying Ding, Technical Report UT-CS-03-504, U. Tennessee, 2003. PRAGMA has more than 30 members around the world, most of them are research centers and universities located around the Pacific. Every member shares at least one cluster with its particular hardware and software specs. This grid environment is characterized by its resource heterogeneity and multiple administrative domains that produce an environment with high levels of variability about its resources availability. Supported By