HPC Advisory Council Kicks Off Fourth Annual RDMA Programming Competition

STEM Students in China Prepare to Demonstrate Programming Prowess

SUNNYVALE, Calif.--()--Focused on supporting STEM student development, the HPC Advisory Council (HPCAC), a leading non-profit organization focused on High Performance Computing (HPC) education and outreach, officially launched its Fourth Annual RDMA Programming Competition in China this week. Designed to support undergraduate curriculum and talent development, this unique hands-on competition furthers students study, experience and mastery.

Launched in 2013, over 20 colleges and universities across China took part in the inaugural competition. Last year’s competition attracted more than 50 colleges and universities drawing attention from world-renowned experts from HPC, cloud computing and Big Data sectors as well as extensive media coverage.

The 2016 competition theme is centered on the world’s supercomputing development, highlighting the theoretical framework, practical model and future development of deep learning, and emphasizes the practical use of competition subjects. Nowadays, “deep learning” has demonstrated powerful development momentum, and how to utilize high performance computing to facilitate deep learning has become an industry-wide focus in recent years.

“Deep learning, big data and data analysis all rely on RDMA technology to improve performance and efficiency, enabling extensive searches in mass data,” said Gilad Shainer, HPC Advisory Council chairman. “This year, we have expanded the scope of the RDMA competition, involving more international universities whose participation brings a new source of vitality and innovation. I believe with the students’ efforts, RDMA’s application in the deep learning sector and the development of next-generation application technologies will be facilitated.”

As an advanced network data transmission mode, RDMA (remote direct memory access) technology has been extensively applied to HPC, deep learning, cloud computing, big data and high-frequency trading sectors. Compared with traditional networks, RDMA can significantly improve efficiency by allowing computers in a network to exchange data directly from memory without involving the processor, cache or the OS of any computer.

“Compared with traditional TCP/IP networks, RDMA features high bandwidth, low latency and low CPU occupancy rate, and is favored by deep learning tasks,” said Dennis Weng, R&D vice president, JD.Com. “RDMA can also significantly improve efficiency in original applications. By virtue of deep learning platforms supporting RDMA technology, an e-commerce platform can, based on user behavior data, rapidly realize ‘user portrait’ analysis, achieve intelligent and efficient mass data processing, and provide users with pertinent and individualized commodities and services.”

This year’s competition will feature an expanded lecture series at universities in major provinces and cities throughout China. Senior RDMA engineers and R&D professionals provide training for students eager to study RDMA and deep learning knowledge. Current planning for the lecture tour includes: Beijing, Shanghai, Guangzhou, Chongqing, Changchun, Xi’an, Harbin, Sichuan, Wuhan, Nanjing and Hefei.

This year’s competition has been honored with the support of multiple institutions. The competition review committee, led by HPCAC chair Gilad Shainer, includes senior experts and leading professionals deep learning including: Dennis Weng, chief technical advisor, JD.Com; Bryan Catanzaro, senior research, Baidu Inc.; Dr. Zhijiang Zhang, iFLYTEK; Dr. Shengen Yan, principal scientist for the algorithm platform, SenseTime; Rick Wagner, HPC system manager, San Diego Supercomputer Center; Pavan Balaji, Argonne National Lab; Tong Liu, director of the HPC Advisory Council, Asia; Rich Graham, MPI forum chair; and Pak Lui, HPC Advisory Council, HPC works special interest group chair.

As high-performance computing, deep learning and big data all continue to expand in both scope and importance, cultivation and selection of student talent has become a priority. The Fourth Annual RDMA Programming Competition encourages students from colleges and universities to participate (as an individual or as a team) culminating with an Award Ceremony during the HPC Advisory Council China Seminar on Oct. 26, 2016. Professors and experts from American universities will grant prizes on the site. One of the many benefits of this competition is the potential for outstanding students to receive offers from elite schools.

In last year’s competition, the National University of Defense Technology and the University of Science and Technology of China won first prize; contestants from Northwestern Polytechnical University and Xidian University were awarded second prize; and the Shanghai Jiao Tong University, Nanjing University and National University of Defense Technology each took third prize.

Students and tutors from the winning teams will be invited to share achievements at HPC Advisory Council international seminars. Tutors will have the chance to explore new research options, strengthen international cooperation and encourage more universities and schools to get involved.

Two first place winners will be awarded 20,000 Yuan each; the two second place winners receiving 10,000 Yuan apiece; and four third place winners receiving 5,000 Yuan each. A team that stands out prominently may also receive a discretionary grand prize award upon unanimous agreement of the expert panel.

Each participating team can consist of one to five undergraduates or postgraduate team members and one tutor. Teams interested in participating should email: rdma@hpcadvisorycouncil.com for more information. Information on the lecture tour: lectures schedule and competition specifics are available at the HPC Advisory Council website. Follow us on our official Weibo @ HPC Advisory Council.

About HPC Advisory Council:

Founded in Silicon Valley, the non-profit HPC Advisory Council is an international organization of over 400 members committed to education and outreach. Members explore high-performance computing technology, application and future development across industry sectors and disciplines. The HPC Advisory Council hosts multiple international conferences and STEM student competitions including the RDMA competition and Student Cluster Competition at the annual ISC conferences. Members contribute to special interest groups and have access to the HPCAC technology center. Membership is free of charge and obligation.

Quotes:

“Nowadays, high-performance computing, deep learning, intelligent voice and other disciplines are seeking more efficient data transmission to improve practical application effects. In particular, deep learning has unceasing demands for computing resources,” noted Dr. Zhiguo Wang, vice dean, iFLYTEK Research Institute. “The RDMA competition provides us with an exciting window into how students use RDMA technologies to optimize programs. Looking forward, we can consider improving network efficiency, releasing system resources and alleviating system burden by means of RDMA communication mode.”

“As a contributor behind the SenseTime face recognition system, RDMA technology can perfectly merge into a traditional data center,” explained Dr. Shengen Yan, principal scientist for the algorithm platform, SenseTime. “While saving CPU resources, RDMA also provides a high-speed low-latency network, accelerates various network services, and helps address the growing technical challenges of computer vision and deep learning. This year’s programming competition is an important platform for selecting the next generation of experts in RDMA. It not only serves as a significant learning experience when students seek employment, but also helps enterprises to cultivate and identify new talent.”

“This year’s RDMA programming competition has evolved into an international event in a very real sense. Top-notch universities from around the world will seek to compete against each other,” predicted Tong Liu, director of the HPC Advisory Council, Asia. “By studying RDMA technologies, college students will be come in close contact with cutting-edge technologies in deep learning and HPC sectors, and have the chance to practice what they have learnt. This can then facilitate closer academy-industry bonding.”

Competition Rules:

Participants: Each participating team consists of one tutor and one to five team members (undergraduates or postgraduates)

Registration method: Send an Email to: rdma@hpcadvisorycouncil.com, indicating name, school, major, grade and contact method of participants, with or without InfiniBand computing cluster, and tutor. We will provide competition test questions and training data, and organize regional training based on registration status.

Competition rules:
1. The Organizing Committee will provide competition source program and training data as soon as an application is received.
2. Applications should be operated in network environment with RDMA capability (e.g. InfiniBand), with operation nodes and configuration subject to the school’s cluster conditions. If RDMA environment is not available to a participating team, the Organizing Committee will provide corresponding operation environment.
3. Participants submit transplanted application source code and execution file, and elaborate the difference in actual operating performance before and after revision and account for performance improvement in a detailed report (English report).
4. The Review Committee reviews source program and reports, and arranges remote or face-to-face defenses, so as to decide final winners.
5. The competition result will be announced at HPC Advisory Council China Seminar (October 26, Xi’an), one of the largest international high-performance computing conferences in China. Several well-known foreign experts will be invited to grant awards to winners at the Seminar.

Competition Sponsor: HPC Advisory Council

Partner Institutions: Cloud Advisory Council, HPCTC, San Diego Supercomputer Center, Argonne National Lab, and China Grid.

Supporting Enterprises (in alphabetical order): iFLYTEK, JD.com, Lenovo, Mellanox, NVIDIA, SenseTime, and Sugon.

Competition Review Committee Chairman
Gilad Shainer, HPC Advisory Council Chairman

Review Committee Members
Rick Wagner HPC System Manager, San Diego Supercomputer Center
Pavan Balaji, Argonne National Lab
Tong Liu, Director of the HPC Advisory Council, Asia
Bryan Catanzaro, Baidu Inc.
Dennis Weng, JD.Com Inc.
Zhijiang Zhang, iFLYTEK
Shengen Yan, SenseTime
Rich Graham, MPI Forum Chair
Pak Lui, HPC Advisory Council HPC Works Special Interest Group Chair
Tarick Bedeir, HPC Advisory Council Programming Advisor

RDMA-related Resources:
http://www.hpcadvisorycouncil.com/pdf/rdma-read-and-write-with-ib-verbs.pdf
http://www.rdmamojo.com
http://www.hpcwire.com/hpcwire/2006-09-15/a_tutorial_of_the_rdma_model-1.html

Contacts

HPC Advisory Council
Cydney Ewald Stevens, 512-770-8916
info@hpcadvisorycouncil.com

Contacts

HPC Advisory Council
Cydney Ewald Stevens, 512-770-8916
info@hpcadvisorycouncil.com