Suzumura Laboratory

Background

Suzumura Laboratory was established in April 2009, headed by Associate Professor Suzumura Toyotaro from IBM Research - Tokyo. Our research focuses on large-scale real-time data processing, stream computing (a major new computing paradigm for high-performance computing), high-speed processing platform, and their applications. For more information related to stream computing, please refer to this page. Laboratory location is on the 9th floor of 8th West Building (W8–905), Ōokayama Campus, Tokyo Institue of Technology. map

Visions and Policies

  • Cooperate with other research institutes and enterprises in exchanging research information for being able to address challenges close to real-world problems.
  • Open culture. Try to broad your perspective on other research area and study research outside of school.
  • Laboratory members should share their new found knowledge, research progress with each other using blog.
  • Promote international awareness by encouraging students to present and write thesis in English.

Latest News (Last updated: October 7th, 2012)

  • A paper titled "Highly Scalable X10-based Agent Simulation Platform and  its Application to Large-scale Traffic Simulation" was accepted to DS-RT conference (2012/10/07)
  • 2 papers were accepted to HiPC 2012 conference. Presentation will be held on December, 2012. (2012/11/08)
  • Suzumura Toyotaro and Koji Ueno were invited to give a keynote presentation at GPU Technology Conference 2012 (2012/10/07)
  • June 2012, won the world's 3rd and 4th places in "Graph500" supercomputer ranking .
  • Two members are going to give presents in May 2012, at IPDPS HPDIC Workshop 2012, Shanghai, China.
  • Performed large-scale graph processing on TSUBAME 2.0 super computer using 1366 nodes at Grand Challenge Program for Graph500 Challenge. (March 28th, 2012) And the result will be announced in International Supercomputing Conference (ISC) 2012.
  • Paper on optimization of Graph500 implementation by Koji Ueno was accepted to HPDC (High-Performance Parallel and Distributed Computing) conference—one of the most competitive conference in high-performance computing area. The conference will be held in Delft, the Netherlands, in June.
  • 3 papers by Koji Ueno, Masaru Ganse, Shunsuke Nishii respectively, were accepted to SACSIS national conference (May 2012, Kobe). For more information of the papers please refer this page.
  • Interview given by Suzumura Toyotaro about stream computing to ITmedia (URL) has been published (2011/12/06)
  • Several press releases about Graph500 result → NEC pageNVIDIA pageMycom Journal, Science and Technology Column on Asahi Newspaper November 24th 2011, Tokyo Institute of Technology Chronicle No. 472 (issued January 2012)
  • Result of Graph500 has been posted here
  • Suzumura Toyotaro gave a presentation titled "Graph500 Benchmark Challenge" at 「第2回シンポジウム-e-サイエンスに向けた革新的アルゴリズム基盤」
  • Won the 3rd  in World ranking of Graph500 Benchmark Challenge using our optimized implementation. For a brief description of Graph500 benchmark, please refer to this page
  • Paper by Suzumura Toyotaro and Koji Ueno was accepted to IEEE IISWC 2012 (IEEE International Symposium on Workload Characterization). (2011/11/07)
  • Paper by Koji Ueno was accepted to Technical Session in NVIDIA GTC Workshop Japan 2011. (2011/07/22)
  • Suzumura Toyotaro gave keynote speech on "Tutorial: Foundation and Practice of Stream Computing" at Tutorial Session of the 28th Annual Meeting of Japan Software and Science Conference.
  • Suzumura Laboratory was featured in Tokyo Institute of Technology magazine named "Landfall". (2011/04/11)
  • Paper by Suzumura Toyotaro and Tomoaki Oik was accepted to IEEE ICWS 2011 in Industry Track (22 papers were accepted, acceptance rate: 11%)  (2011/07).
  • Paper by Atsushi Ishii and Suzumura Toyotaro was accepted to IEEE Cloud 2011 Research Track. (acceptance rate: 18% = 36/196) (2011/07)
  • Paper by Shunsuke Nishii and Suzumura Toyotaro was accepted to SACSIS 2011.
  • Suzumura Laboratory was awarded the IBM SUR (Shared University Research) Award (2010/11/01). → Award page
  • Paper by Koji Ueno and Suzumura Toyotaro was accepted to Internet Conference 2010 and received Student Encouragement Award - (2010/10/26)  → Award page
  • Paper by Shunsuke Nishii and Suzumura was accepted to Internet Conference 2010 - (2010/10/26).
  • Paper by  Atsushi Ishii and Suzumura was accepted to Internet Conference 2010 in WIP session - (2010/10/12)
  • Paper by  Hiroya Matsuura,  Masaru Ganse and  Suzumura Toyotaro was accepted to COMSYS 2010 (translated: System Software and Operating System Conference) - (2010/10/12)
  • Paper by Kosuke Morita won the Excellence Award at the 72nd National Convention of the Information Processing Society of Japan
  • Paper titled「ストリーム処理 System S を基盤にした Web サーバーの実装」(translated: Implemtation of Web Server Based on System S) by Tomoaki Oiki and Suzumura Toyotaro, presented at Internet Architecture Conference (IA) in 2009, received IA Research (2010/05).  →  IA Award page
  • Paper titled 「バッチ処理とストリーム処理の融合」(translated: Integration of Stream Processing and Batch Processing)  by Hiroya Matsuura and Suzumura Toyotaro received Student Encouragement Award at National Convention of the Information Processing Society of Japan. (2010/03)

About Stream Computing

Our research focuses on new computing paradigm namely stream computing. Stream computing can be simply described as continuous computing or querying on data stream that arrives in real time and in large quantity. Traditional data processing is based on store-then-process paradigm, however, in recent years data is growing explosively, so analyzing data using traditional approach will take very long time and require enormous storage capacity. Stream computing is the answer for this problem.

The concept of stream computing can be described as follows: arriving data is kept in buffer on memory, then it is passed sequentially to processes to perform data analysis without storing to storage. This helps minimizing enlargement of storage and ensuring the analysis done in real time. Some research area such as digital signal processing (DSP) has also used this concept for long time, however, it supports data from specific domain. Streaming Computing is not domain specific, it is a generic software processing technology which is suitable for processing various type of data (structured and unstructured) from various source with large data rate. for that reason, many research themes related to stream computing are of interest such as programming model, scheduling techniques, runtime optimization. According to ICDE/SIGMOD/VLDB, which are famous conferences for database system, stream computing is a hot research topic in recent years. The following is a summary of research themes of Suzumura Laboratory as of October 2009. 

Please click at the picture to enlarge.