This past week we worked on getting Aprof to function properly on MareNostrum. We amended Aprof’s source code to fix the way that it inputs data. We amended Amon’s source code to account for the different version of Linux that is installed on MareNostrum in comparison to GCB and Mind. We developed a script to combine the Amon outputs from each node into one file and to properly configure the data into a Aprof readable format. We, also, worked to generate predictions using Dimemas and received a demonstration from the Dimemas developer. Personally, I worked extensively on my 2nd REU Progress presentation and a bit on my Linux studies as time allowed.
In this part, please only mention your significant accomplishments that you have not reported previously. Note that you do not have to have an accomplishment each week.
-Aprof source code successfully amended
-Aprof functioning properly-Amon source code successfully amended
-WRF v. 2.2 compiled on Mind
In this part, please mention the minor and major problems that you are facing in your research.
-We are still faced with some decisions on what we are going to do about the different versions of WRF installed at the different sites, GCB, Mind, and MareNostrum. This is an important issue if we want to be able to make comparisons between the results generated on each of the different clusters. Right now v. 2.2 is installed on Mind and GCB and v.2.2.1 is installed on MareNostrum
In this part, please mention your short- and long-term plans for the weeks and months to com.
At this point we have obtained objectives 1, 3a, and 3b, and we are continuing to work on objectives 2 & 4. As this is our final week, we have set out two special objectives to complete before going home or shortly thereafter.
1. Get Aprof query automation and comparison script up and functioning properly on MareNostrum. Obtain a results set of Aprof predictions for the different scenarios we are simulating for.
2. Have full knowledge of how to run Dimemas simulations and set up Dimemas simulations so that the project can get its first look at comparison of the two tools.
These are our main objectives to close out our experience here at the BSC. Once these are complete, we can reassess the specific goal of the research with the first results and prepare to arrive at our significant conclusions by the end of the summer. With this we will keep in mind the prediction accuracy of Aprof and Dimemas and how they compare and contrast to each other. This should allow us to improve our approach and present our significant follow-up findings.
Personally, I will continue my studies in the Linux platform.
Barcelona Research Objectives
Overall Objective: Achieve accurate and timely performance prediction on compute cluster to be used in Grid Computing Environment Meta-Scheduling.
1. Get mpidtrace linking properly with WRF compiled on GCB, then Mind. GCB COMPLETED
2. Use generated MPI tracefiles (Paraver and Dimemas) to do prediction between Mind and GCB IN PROGRESS
3. a) Install Amon and Aprof on MareNostrum. COMPLETED
b) Run benchmarks on MareNostrum FIRST RUNS COMPLETED
4. Experiment with how well Amon and Aprof scale to larger number of nodes IN PROGRESS
5. Analyze how Amon and Aprof relate to/could possibly combined with Dimemas.
6. Work with Marc to see how we can optimize the gridification of WRF
SUMMARIES/CRITIQUES OF PAPERS:
Readings temporarily satisfied.