PPL at Illinois Emerges as a Powerhouse in Parallel Computing at SC14
The Parallel Programming Laboratory (PPL) at the University of Illinois at Urbana Champaign is a research group that has been creating innovative solutions in HPC for over 25 years. Adaptive runtime systems is the core research focus of PPL. The research results are embodied in the Charm++ parallel programming system and the Adaptive MPI library, which have been used to develop many highly scalable applications, including NAMD (biophysics), OpenAtom (materials), ChaNGa (astronomy), EpiSimdemics (simulation of contagion spread), etc.
Charm++ is a strong contender for an exascale programming model, with all the adaptive features built into Charm++, including dynamic load balancing, power/energy/temperature management, and resilience.
The Supercomputing 2014: International Conference for High Performance Computing, Networking, Storage and Analysis (SC14) will be in New Orleans, Louisiana mid-November. Again this year, the Parallel Programming Laboratory will be well-represented starting with a half-day tutorial on Resilience and then another half-day tutorial on Charm++. The conference’s technical program will include 6 papers by PPLers and collaborators. That’s not all! Harshitha Menon will receive a George Michael Memorial High Performance Computing Fellowship. Akhil Langer and Ehsan Totoni will give presentations on their research during the conference’s Doctoral Showcase. Xiang Ni and Ehsan Totoni will be presenting in the ACM poster competition. PPL will host a Birds-of-a-Feather session. The complete schedule can be found here.
PPL Director, Professor Kalé, and Kathryn Mohror, research scientist at LLNL, will teach a tutorial titled " Practical Fault Tolerance on Today's Supercomputing Systems." This tutorial will present introductory material for developers who wish to learn fault tolerant techniques available on today’s systems. We will give background information on the kinds of faults occurring on today’s systems and trends we expect going forward. Following this, we will give detailed information on several fault tolerant approaches and how to incorporate them into applications. Our focus will be on scalable checkpoint and restart mechanisms and programming model-based approaches. This tutorial will run from 1:30PM – 5:00PM on Sunday, November 16th.
Professor Kalé and PhD candidates, Nikhil Jain and Michael Robson, will teach a tutorial titled "Parallel Programming with Migratable Objects for Performance and Productivity." This tutorial will focus on application development following the principles of migratable, message-driven objects that underlies the Charm++ parallel programming environment. Attendees of the course will learn how to construct next-generation parallel applications in this setting in order to benefit from automatic load balancing, fault tolerance, and easy and efficient parallel composition. This tutorial will run from 8:30AM – 12:00PM on Monday, November 17th.
The 6 papers are:
- Mapping to Irregular Torus Topologies and Other Techniques for Petascale Biomolecular Simulation
- Maximizing Network Throughput on the Dragonfly Interconnect
- Parallel Programming with Migratable Objects: Charm++ in Practice
- Maximizing Throughput of Overprovisioned HPC Data Centers Under a Strict Power Budget
- Optimizing Data Locality for Fork/Join Programs Using Constrained Work Stealing
- Using an Adaptive HPC Runtime System to Reconfigure the Cache Hierarchy
James Phillips | Yanhua Sun | Nikhil Jain | Eric Bohm | Laxmikant Kale
Nikhil Jain | Abhinav Bhatele | Xiang Ni | Nicholas Wright | Laxmikant Kale
Bilge Acun | Abhishek Gupta | Nikhil Jain | Akhil Langer | Harshitha Menon | Eric Mikida | Xiang Ni | Michael Robson | Yanhua Sun | Ehsan Totoni | Lukasz Wesolowski | Laxmikant Kale
Osman Sarood | Akhil Langer | Abhishek Gupta | Laxmikant Kale
Jonathan Lifflander | Sriram Krishnamoorthy | Laxmikant Kale
Ehsan Totoni | Josep Torrellas | Laxmikant Kale
PhD candidate, Harshitha Menon, will receive a George Michael Memorial High Performance Computing Fellowship for her project "Scalable Load Balancing and Adaptive Run Time Techniques." The awards ceremony will be at 12:30PM on Thursday, November 20th.
For the Doctoral Showcase, there were 44 submissions and only 16 were selected. 2 PPL PhD candidates will present:
- Akhil Langer is focused on "Parallel Algorithms For Two-Stage Stochastic Optimizations" and will present at 3:30PM on Thursday, November 20th.
- Ehsan Totoni is focused on "Adaptive Power Efficiency: Runtime System Approach With Hardware Support" and will present at 2:45PM on Thursday, November 20th.
PPL will host a Birds-of-a-Feather Session on "Charm++: Adaptive and Asynchronous Parallel Programming." This is an ideal opportunity to learn more about how Charm++ is used in practice, see how it's growing, and influence its future direction. The session will take place from 12:15PM – 1:15PM on Thursday, November 20th.
Many PPL members will be in attendance at the conference. We look forward to an exciting week at SC14.
At the PCI booth, PPL members will show Charm++ Fault Tolerance demo. In this demo, we will show the online failure handling using Charm++. After killing one node by powering it off, you could see the visualization of the demo program pauses due to restart and quickly proceeds from the last checkpoint.
List of PPL Activities at SC14:
PPL is showcasing its success and talents in what proves to be its best year yet. This is an encore after a very successful SC13 performance.
- PPL members have six papers in the main conference technical program. Probably the highest for any research group. At SC13, PPL members had 4 papers in the main conference technical program.
- PPL PhD Candidate Harshitha Menon won the prestigious George Michael fellowship, there will be an awards ceremony at 12:30PM on Thursday. Harshitha was also a best student paper finalist at SC13. Also at SC13, PPL student Jonathan Lifflander won the George Michael fellowship, while the only other winner (Edgar Solomnik) was a PPL alum! Yanhua Sun received honorable mention for the George Michael fellowship at SC12.
- A recent PPL alum, Abhinav Bhatele will receive the IEEE TCSC Young Achievers in Scalable Computing award at SC14.
- PPL will
- Host a Charm++ BoF (Thursday noon)
- Present a tutorial on Charm++ (Monday 8:30AM)
- Co-present another tutorial on resilience (Sunday 1:30PM)
- Show Charm++ Fault Tolerance Demo (Parallel Computing Institute booth)