Speaker
Description
URL for further information
http://grb.spaci.it/grb/, https://sara.unile.it/libi/, http://www.libi.it
Conclusions and Future Work
We present a first prototype of a workflow engine that brings the power of heterogeneous Computational Grids to the desktop, and allows the users to conveniently assemble and run their own scientific workflows. Future work will regard: i) the full implementation of the Meta Scheduler, regarding data tasks support, and related testbed on real bioinformatics case studies; ii) the development of a plug-in to support the CREAM service of gLite, and iii) the development of the job monitoring service.
Keywords
Grid Workflow Management System, Meta Scheduler, gLite, Unicore, Globus
Detailed analysis
The main goal of our work is providing an integrated environment for interoperability among different grid middleware such as gLite, Unicore and Globus. It allows the composition of jobs (batch, parameter sweep and MPI based) and data tasks exploiting the GRelC Data Access Service (DAS). The engine implements the logic to execute them by using a standard language such as JSDL, OGF compliant, that has been extended for this purpose. Our workflow engine has been designed taking into account the following requirements: (i) ability to handle workflows described by DAGs; (ii) ability to handle complex workflows described by arbitrary graphs, supporting cycles and conditions; (iii) support for recursive composition, i.e. the possibility to define a workflow vertex as a sub-workflow or parameter sweep vertex instead of a batch task; (iv) support for secure remote execution by means of grid security infrastructure (GSI) protocol.
Impact
In a first version our workflow was based on Globus. We have considered also gLite and Unicore as requirement of the LIBI project since involved computational resources are based on several heterogeneous middleware. An initial comparison of our system with several tools such as P-Grade and GridWay and several bioinformatics workflows such as Taverna, Pegasys and Bio-Steer has shown that they do not fully satisfy the requirements of interoperability with Unicore and several features such as the support for cycles and conditions, needed for example in the bioinformatics experiments. Moreover, the ability to specify data tasks, represents an important feature of our system.
Our system is interoperable with the Workload Management System (WMS) and Logging & Bookkeeping (L&B) services of gLite, with the Network Job Supervisor (NJS) service of Unicore (through the Gateway), and with the GRB scheduler for the Globus Toolkit, by using the JSDL standard language.