 |
||
| The map above shows current traffic flows and traffic patterns across GLORIAD - as measured in Chicago. In the map, triangles show source of traffic; inverted trianges, destination. Sites connected by arcs are the heavist current consumers. Colors indicate network protocols and can be discovered by the listing of top users at the right. The small charts show traffic volume by application during the last two hours. Clicking on any of them takes you to a larger display. | ||
Project: Genome Analysis and Databases Upload/Gnare
Description: The GADU application automatices the process of function assignment to gens in genom analysis. It works by finding new genomes in public databases. Each of these genomes is composed of ~4k genes which need to be processed and characterized. These characterizations are then stored for future use.
Participants: Computational Biology Group at ANL http://www-unix.mcs.anl.gov/compbio/
Sponsors: DOE Scientific Discovery through Advanced Computing (SciDAC)
Countries Involved: US
Tools: Prior to using Globus, GADU used custom resource-specific scripts for job submission and control making it difficult for them to take advantage of new or existing resources as they became available. The resource of choice was the ANL Chiba city cluster and the turn around time for the processing of a single genome was ~5 hours. GADU was modified to use Grid tools for it job submission and control processes, namely the Globus Toolkit for data movement and job submission, Condor-G to manage the job stream, Chimera for data flow control and Java CoG kit. This standardization allowed it to be independant of custom resource-specific job mechanisms. This allowed for a run of 59 genomes from DOE-sponsored sequencing projects in 24 hours by using resources available on the DOE Science Grid (ANL and PNNL). This represented 67 CPU-days of processing time, ~10k jobs, 50 GB of generated data and over a 5 fold improvement in turnaround time (~5 hours/genome to <1 hour).
Contact: Alex Rodriguez
