I am currently working in a research group that focuses on scientific workflow management in distributed environments. The main project in this group is the Pegasus workflow management system (http://pegasus.isi.edu). Pegasus was used as part of several NSF Information Technology Research projects: Grid Physics Network (GriPhyN), National Virtual Observatory, the Southern California Earthquake Center (SCEC) Community Modeling Environment, and others.  Using Pegasus, earthquake scientists are able to generate more accurate hazard maps that can be used by civil engineers to design new construction in earthquake-prone areas. Astronomers use Pegasus to generate large-scale (6 and 10 square degree), science-grade mosaics of the sky that allow them to see structures not observed before. Gravitational-wave physicists are using Pegasus to run sophisticated analysis in the hopes of finding gravitational waves. Neuroscientists analyze 3D images of the brain in the hopes of understanding complex brain functions.


I am also exploring the new “cloud” technologies and how they can be used as an execution environment for a workflow system. As part of this work I will develop an interface between the Pegasus workflow management system and the cloud resources and characterize the performance of the workflow execution in such an environment.


I will be learning about workflow technologies and the challenge of managing data and computations on distributed resources. I will also help define a workflow for a new application from the bioinformatics domain. This will include familiarizing myself with the Pegasus software and the input format it needs as well as the understanding of how the application is set up and how it needs to be represented as a workflow.



I have immersed myself in Grid technology and the Pegasus project by reading as much as I possibly could about them.  Just when I thought I had read all I could find, my mentor sent me links to more papers. I will start work on a project next week dealing with sequencing genomes.



The  genome project is being put on hold until we can find a time to meet with the biologists.  In the meantime, I will be trying to run workflows on the Nimbus Cloud (http://workspace.globus.org/clouds/) for the Montage astronomy application which delivers science-grade mosaics of the sky.  I have been doing a lot more reading, and I have downloaded and installed the workspace cloud client from Nimbus, as well as the Pegasus software.


I have fired up my first virtual machine, but I am having trouble getting workflows to run.  This week I have installed GridFTP from the Globus toolkit, GRAM pre-WMS server, Condor, xinetd, and the Pegasus worker packages on a virtual machine (VM) created from the image Globus-002 (from the cloud client workspace).  The virtual image Globus-001 did not have enough space, so after consulting with Tim Freeman of the University of Chicago, he created a new one that would suit more needs.  Each VM runs for a finite amount of time, specified by the user at deployment.  A copy of the VM image must be saved before termination if any changes have been made.  A simple certificate authority was also created on the virtual machine this week.  


I tried to run the Pegasus tutorial that uses a simple Montage workflow on one virtual machine, but ran into problems with the format of the site catalog, replica catalog, and transformation catalog that Pegasus uses to plan workflows, as well as with the setup of a cluster.  I had the tutorial running at the end of the week on one, two, three, and four virtual machines, so I should be able to start running workflows for Montage next week.


I was able to get some of the larger workflows running on a virtual machine this week, but for some reason the largest one of them will not work.  I think it may just be too large for the size of virtual machine image I am using.


I was able to make some good progress with the science cloud this week.  I have run four different sized workflows on 1, 2, 3, and 4 virtual machines, and will begin running the same workflows on the local cluster next week.


A shorter week because I had to fly home for a wedding, but I still got a lot done.  I was able to run the workflows on the local cluster and on my local machine as well.  I have gotten some initial numbers, and will begin writing a paper to submit to a workshop at the e-Science conference when I get back.


I have started experimenting with a virtual cluster to make better comparisons with the local cluster, and I also have a good start on my paper.  The virtual cluster is interesting because the virtual machines must become aware of each other so that they can work together.   I have been unsuccessful in getting my workflows to run on the virtual cluster so far, but it looks promising.


With one week left, I am starting to feel quite a time crunch to get this paper finished.  The virtual cluster still is not working, and my results are not making a very good comparison.


I got the virtual cluster working and was able to do a lot more runs to get better numbers for the paper.  Now that I am home, all I have left to do is editing before I wait to see if my paper gets accepted.


I finished the paper, and now I am waiting to hear whether it got accepted to the SWBES workshop.  Here is a link to my draft of the paper.