Return 2012 MSM Working Group discussions
Discussion Topics
1. Do you need access to computing resources that are not currently available to your multi-scale modeling needs?
2. If you don’t have access to such computing resource, we will arrange webnars to introduce the following topics (expected in Nov – Dec 2012)
- a) Manage your files and use the high performance computing system
- b) Introduce parallel processing.
3. GPUs and other non-standard computer hardware is being used to solve the multi-scale problems. Would you benefits (and also willing) share your experiences and code to benefit members who do not have expertise or human resources to develop the code on their own?
4. Would cloud computing benefit your investigations? Would you like to access cloud environment such Amazon EC2?
Please give us your comment and topics you want to include so we can better prepare the webnars.
Madhav Marathe:
- Role of HPC in developing scalable models
- a broader definition of HPC that goes beyond computing systems to associated software systems as well
MSM High Performance Computing Working Group Discussion Memo
Five people including a program manager from DOE attended the high performance computing working group (HPCWG) discussion. During the discussion, we went through the topics listed in the HPCWG Wiki page, reached some agreements, and proposed possible activities in the year of 2012-2013.
Topic 1: Computing Hardware Resources
All attendees claimed that they had some access to high performance computing resources and HPC resources would benefit their research. Due to different research interest, the attendees expressed different requirements on the hardware. Some group member requires >200 nodes to perform their research. This requirement is really hard for most of the high performance computing centers. To address this issue, we all hope that we can group together to apply for large space usage and sharing nodes within the MSM consortium. Since a program manager from DOE also attend the discussion and indicate some interest in working with the HPCWG, we will follow up to contact with DOE for possible help. No group member expressed any interest in GPU.
Topic 2: Computational Algorithms
All attendees agreed with the importance of sharing developed tools for high performance computing. Specifically, current available freeware and shareware that help them to build multi-scale model benefit them most. All attendees agree with sharing these tools through HPCWG and email list. Since HPCWG members have different working experiences with HPC, we didn’t reach a wish-list of optimized libraries for the HPC environment. In order to improve the working efficiency, we agreed that we will organize a webinar to demonstrate how to use HPC with a small project. This will be followed up by the group leader early by March 2013.
Topic 3: Collaborate, define the needs for the community
We have formed the following collaborations and defined the need for the community for the 2012-2013 year. 1. Try to apply some shared node and storage spaces for the HPCWG. Since most of us still want to use free resources, we expect some difficulties when we apply for this. 2. We will share useful tools, algorithms, and software in the HPCWG. 3. The HPC WG will organize two-three webinars in the following year: a) how to use current HPC resources; b) How to program for parallel computing to improve research efficiency; c) Demonstrate the usage of HPC with a small project. (These webinars might be combined together.)
In summary, the discussion provided a good opportunity for the HPCWG members to know each other and formed some agreements for the working group to collaborate with each other. Specifically, it clearly defined the needs of the community, thus, established the aims for the WG to work on this year.