Life and Career Development: a Perspective from a Chinese Scholar Haibo Chen Shanghai Jiao Tong University http://ipads.se.sjtu.edu.cn/haibo_chen
Life and Career Development: a Perspective from a Chinese Scholar Haibo Chen Shanghai Jiao Tong University http://ipads.se.sjtu.edu.cn/haibo_chen
Outline • Self-introduction • Growing with the Academic Community • Experiences in Academia and Industry • Interaction b/w Academia and Industry • Mind the Gap: Academia vs. Industry • Some Experiences
Self-introduction • Ph.D, Jan 2009 from Fudan University • Assistant Professor at Fudan since April 2009. • Professor at Shanghai Jiao Tong University since December 2011 • Research group: Institute of Parallel and Distributed Systems (IPADS): http://ipads.se.sjtu.edu.cn • Director and Chief Scientist for OS since December 2016 (sabbatical position) • Operating System Kernel Lab@Huawei • Research Area • Computer systems in general: OS, Virtualization, In- memory computing systems, system security, etc.
Grow with the Community: Engagement w/ChinaSys and APSys AsiaSys APSys ChinaSys
9 Years Ago (Nov. 2008)
APSys: Asia-pacific Workshop on Systems • Started from 2010 • 2010 at Indian: GC: Roma Kolta, PC Chair: Lidong • 2011 at Shanghai: GC: ZZ and I, PC Chair: YY and Sue Moon • 2012 at Korea: GC: Sue Moon, PC Chair: Satya • 2013 at Singapore: GC: Ben Leong, PC Chair: Junfeng and ZZ • 2014 at Beijing: GC: Yungang and Wenguang, PC Chair: Lorenzo and I • 2015 at Japan: GC: Kenji Kono, PC Chair: Robbert and Kyoungsoo • 2016 at HongKong China, GC: Heming and Francis, PC Chair: Sorav and Lin • 2017 at Mumbai, India, GC: Puru, PC Chair: Prashant and Lintao
The Story of ChinaSys • Online discussion began during organizing APSys 2011 (prior code name: DeepSea) • Initial idea by ZZ, and make it real together with Wenguang, Lidong and Haibo • Founding discussion right after the banquet of APSys • ZZ, Wenguang, Binyu, Xiaofei, Yingwei, Yinhe, Yu, Yungang, Xiaolin, Haibo, etc… • Inaugural workshop: Chair: Wenguang, Shenzhen after CNCC • 2 nd : Chair: Haibo, Hangzhou, 3 rd : Chair: Xiaofei, Wuhan • 4 th : Chair: Yungang&Yinhe, Beijing, 5 th : Chair: Yu, Suzhou • 6 th : Chair: Li, Changsha, 7 th : Chair: Yingwei, Shenzhe • 8 th : Chair: Zhibin, Xiamen, 9 th : Chair: Haibo/Weihua, Shanghai • 10 th : Chair: Xiaofei, Wuhan, 11 th : Chair: Yungang, Shanghai Becomes the third SIGOPS chapter since August 2015
Original Goal • Form a “club” with faculties from several research groups, invitation only. • The faculties are the young working horses, instead of already established. • Every half year, one of the club member will be the host of the 1day event. • Faculties take their senior students, and MSRA research team (in most cases the entire team) will join for a one day retreat. • Opportunistically, we will invite one external senior researchers(not necessarily from any of the members). • The goal is brainstorm and plan on works that will push this small community to reach for an ambitious and yet achievable milestones. • For the system discipline, it would something likeSOSP/OSDI/ASPLOS papers. Ideally, this is to promote collaboration among the members
We Are Growing and Stronger Since! • Numerous papers published in top systems conferences like SOSP/OSDI, EuroSys, Usenix ATC, and top architecture conferences like ISCA, MICRO, HPCA, ASPLOS • 2015 (Incomplete) • 1/30 SOSP paper • 3/32 EuroSys papers (1 best award) • 5/47 ATC papers • 3/23 PPoPP papers • 5/58 ISCA papers • 2/48 ASPLOS papers • Others: 2/29 MobiSys, 2/60 Usenix Security, • Several best paper awards/nominee: ASPLOS, MICRO, HPCA, EuroSys, APSys
Community Services: Help You, Help Me! • 2017 ACM SOSP General Chair (w/ Lidong Zhou@MSR) • ACM APSys Steering Committee Chair (w/ Lorenzo Alvisi@Cornell) • ChinaSys Co-founder (2011), ACM SIGOPS ChinaSys Vice Chair (2015) • PC: SOSP , EuroSys, Usenix ATC, ISCA, ASPLOS, Oakland, PPoPP , CCS • Associate Editor: ACM Transactions on Storage
Outline • Self-introduction • Growing with the Academic Community • Experiences in Academia and Industry • Interaction b/w Academia and Industry • Mind the Gap: Academia vs. Industry • Some Experiences
Academia vs. Industry Industry: Academia: Innovation & Platform & Incubation Product
Difference b/w Academia and Industry Paper (+prototype) vs. Product P2P vs. Hierarchical Hypothesis-oriented vs. market-driven Free vs. well-organized Scientific insights vs. market insights
Difference b/w Academia and Industry Time constraint: paper deadline vs. time to market Impact: knowledge vs. product Persistent vs. fast changing Communication/interpersonal skills
Mind the Gap: Be Careful of Flexibility We build stuffs with flexibility Reinvent a wheel! Invent an irrelevant wheel!
Mind the Gap: Eco-system Matters • It is usually hard to destroy an existing eco- system • High risk research: find chances for a revolution • Low risk research: fix critical issues when evolving existing eco-system • Key: needs an evolutionary path
Mind the Gap: Last Mile is Really Long • Belief: I build a good system with innovative ideas, industry will definitely adopt it into its product • Reality: industry cares a lot on the maturity of a system due to pressure on TTM, you usually have some competitors • Suggestion: do last mile work yourself if you really want to push work into industry
Mind the Gap: Last Mile is Really Long • A long way to go from a research prototype to a product • A research paper only finish ~1/10 of the journey • Lots of traps ahead from research prototype to a product • Compatibility issues • Reliability issues • Deployment efforts • Dirty work to hacking • Need a strong incentive to do so: 10-100X performance improvement, a revolutionizing security foundation
Mind the Gap: Complexity • Computer systems are already very complex • Academia: usually needs to make some simplified assumptions to conduct research • Industry: 10+X complexity when applying a research idea back
Tip: Gaining Insight on the Technical Trend • Make sure it aligns with trend with industry • Industry: 1-2 years ahead for product line, 2-5 years ahead for research lab • Academia: 3-10+ years ahead • Constraints: Time to Market
Tip: A Typical Way of Research: Incommensurate Scaling As a system increases in size or speed, not all parts of it (&human factors) follow the same scaling rules -- Saltzer & Kaashoek X 10 million = Key: skeletal system 22
Tip: Incommensurate Scaling As a system increases in size or speed, not all parts of it (and human factors) follow the same scaling rules (Saltzer and Kaashoek ) Examples: Memory wall: CPU speed increases faster Multicore scalability: #core increases faster Mobile security: security not match open infrastructure In-memory computing: data access > 1,000X faster
But the Gap Considered Useful • A gap helps distinguish academia and industry • Academia: cutting-edge research with great flexibility • Industry: comprehensive full solution, customer-oriented
Summary • Grow with the community: help you, help me! • From academia to industry • Mind the gap • Cope with fast changing world
Thanks
Recommend
More recommend