AoE Proposal:Institute of Network Coding

August 25, 2009*

Conventional Transmission Paradigm ParcelsTransportation networkCargo shipment*

ParcelsTransportation networkCargo shipmentInside the network: Store-and-forwardConventional Transmission Paradigm *

Telecommunications inheritedStore-and-forward from the transportation networkInternet*

Information Parcels, because it can be processed. *

Information Parcels, because it can be processed. *

Information Parcels, because it can be processed. Network Codingbetter thanstore-and-forward*

Application example: Space communicationsABAB*

Application example: Space communicationsABABAA+BB*

Application example: Space communicationsABABAA+BBSame model applies to satellite and wireless communicationsPrototyped by MIT on the wireless LAN (Wi-Fi, 802.11)*

NC for Data StorageDisk A Copy ADisk C Copy CCopy B Disk BCopy D Disk D *

Disk ADisk CDisk BDisk D A single backup disk ABCDNC for Data Storage*

NC for File-sharing (P2P)*

More applications of NCInternet telephonyInternet TVsSensor networks*

Performance improvements by NCBetter receptionClearer imagesMore privacy*

Improvements everywhere*

Special Report in Scientific American

Breaking network logjams, Scientific American, June 2007Although this method may sound counterintuitive, network coding, which is still under study, has the potential to dramatically speed up and improve the reliability of all manner of communications systems and may well spark the next revolution in the field.*

NC is a New ParadigmFundamentally changes the concept of network communicationWith NC, a new information superhighway system

*

Tremendous impactInformation theoryComputer networksSwitchingtheoryWireless networksChannelcodingComputerscienceCryptographyIntracellularCommunications OptimizationtheoryQuantum information theoryGraphtheoryGame theoryMatroid theory*

Research impact of NCANNUAL CONFERENCES (2)*WiNC

Research impact of NCCOURSES IN TOPUNIVERSITIESRESEARCH LABSIndustry/Defense*

NC in the News(2006)HK Newspapers(2005)(2007)(2008)*

The research team (of 23) 5 IEEE Fellows13 best paper awards69 US patentsPrincipal Investigators*

Local academic Co-InvestigatorsMinghua ChenDah-ming Chiu Xiaodan FanSidharth JaggiLap Chi LauJack LeeSoung-Chang LiewJohn LuiChandra NairPing LiAlbert SungGuangyue HanWai-Ho Mow*

Non-local academic Co-InvestigatorsZhen ZhangNing CaiYixian YangHenry LuJohn Zao*

Yixian YangIndustrial Co-InvestigatorsDr. Philip ChouPrincipal ResearcherDr. Mingxi FanDirector of EngineeringDr. Douglas ChanSenior Wireless Systems Engineer*

NC founded at CUHK2005 IEEE Information Theory Society Paper AwardNC founded at CUHK in late 1990s.

*

Our recent effortsChannel coding theoryComputer networksCryptographySwitching theoryNetwork codingModern theoryof communicationSignal NCImproved wirelesscommunications*

Why AoE in HK?Very large research field*

Why AoE in HK?Very large research fieldUS/Europe Research conglomeratesHuge military projects*

Why AoE in HK?Very large research fieldUS/Europe Research conglomeratesHuge military projectsNeed critical massInstitute of Network Coding*

Unique strength of this AoEAcademicStrongest theory group in the worldWide expert areasIndustryThree technology giants*Support a complete knowledge supply chain

Unique advantages of Hong KongNumerous researchers visit HKTelecom industry in Shenzhen (e.g., Huawei, ZTE)Help China, Mainland and Taiwan, to catch up in NC

*

Goals of the AoENetwork Coding*

Goals of the AoECover all existing areas and steer the field

Network Coding*

Goals of the AoEHost international conferencesNetwork Coding*

Goals of the AoEGain global recognition in researcher trainingNetwork Coding*

Goals of the AoEConduct technology consulting for HK, Mainland China, Network Coding*

Goals of the AoEHelp attract major industrial research labs to HKNetwork Coding*

Now is good timingThe Network Coding Research Centre (since 2006, CUHK) in 2006 paves for the AoE.Jump-started NC research in sister universities.Held NetCod 2008 on CUHK campus.Elevated research interest in the region.Funding > $6M (CRF, NSFC-RGC, industry, others)Many good PhDs in NC are being produced when job market is bad.Acting now, we shall be the first Institute of NC.*

DeliverablesTrainConferences15 PhDs40 PostdocsHong Kong 2013/15Beijing 2011Hong Kong 201xInstitute of Network CodingWiNC*

DeliverablesInstitute of Network Coding*

Application ProjectsPotential for significantimpact within 5-10 yearsWireless NC (with Qualcomm)Re-engineering Internet backbone (with Cisco)Security, reliability for File-sharingDisk A Copy ABandwidth, energyoptimization forData storage systemsPattern discovery for Biological sequencesNext-generationInternet Protocol TV (with BUPT/ZTE)Scalability forVideo coding

*

Interconnecting test-bedsWireless Software Radio LANRunning in 2 yearsQuick evaluation of NC protocolsPerhaps also campus-wide wireless test-bed*

Budget (15 local Co-Is, 8 years)26.9123.453.36(Million HKD)Total 103.67 MBudget breakdown*

Staff4 ResearchAssistantProfessorsIndependent research in NCKeeps track of the field4 Visiting ResearchFellowsEminent researchers in related fieldsWiden team members exposureBoost Institutes image2 EngineersBuild systems/prototypes10 PostdoctoralResearchersSupervised by 15 local PIs/Co-Is*

Governance structure*

GovernanceExecutive Board of GovernorsFiscal mattersVC, Bursar (CUHK), PCScience & Engineering mattersLong-term visionPromote visibility3-5 prominent scholars Prof. Andrew YaoTuring award winner*

GovernanceExecutive Board of Governors(Meet annually)The two PIs

Prof. Bob LiProf. Raymond Yeung

Prof. Vincent ChanMIT

Dr. Nim-Kwan CheungCEO, ASTRI (HK) 3 prominent academic/industry membersSet/review directionsOverall budgetAd-hoc committees*

Governance Advisory BoardScientific Advisory BoardResearch CommitteeAnnual/quarterly meetingsExecutes BoG decisionsExecutive Board of Governors(Meet annually)All PIs/Co-PIs (23)Proposes & consolidatesprojects*

Sustainability scenariosAoEMission accomplished All NC ideas developed and productizedSteady progressContinuous productizationFairly self-sustaining*

ConclusionNeed an AoE to sustain our global leadership.Bring substantial benefit to Hong KongResearch, education, industry, economyMake the Hong Kong success story even bigger.*

Thank you*

*Hello, I am Raymond Yeung from CUHK. I am a P-I and the Project Coordinator of this AOE proposal.In this presentation, I will introduce to you a new paradigm in Information Science called Network Coding. This new paradigm was invented at CUHK.Conventional network communication is very much like delivering parcels through a transportation network. Inside the transportation network, we store and forward parcels, just like in the postal service. For decades, this store-and-forward paradigm has been inherited by communication networks. In other words, information is treated like parcels. Folklore for network communications says that treating information like parcels is good enough. However, instead of just store-and-forward, the network can actually process information. For example, the network can copy information, can combine information, can transform information In the late 1990s, I started to explore this freedom in information delivery that eventually led to a major breakthrough in information science and beyond. Together with my collaborators, including P-I Bob Li sitting here and Co-I Ning Cai in this picture, I proposed a new paradigm for network communications. This new paradigm is called Network Coding.

And surprisingly, we discovered that Network Coding can significantly improve over store-and-forward. This seminal work of ours marked a new era in network communications.While Network Coding is a highly sophisticated mathematical theory, its advantages can be illustrated by this simple example taken from a recent proposal for a NASA project.

Suppose two rovers on Mars want to exchange messages A and B through a satellite; the yellow message A going from left to right, and the blue message B from right to left.

In the store-and-forward paradigm, after receiving the two messages, the satellite needs to transmit both the yellow and the blue messages individually. In other words, the satellite needs to transmit twice.On the right is the Network Coding scheme. Here, after receiving receives both messages A and B, the satellite combines them into a new message A+B, and then broadcasts this new message to both rovers simultaneously.

The rover on the left already knows its own message A. Upon receiving the message A+B as side-information, it can figure out the message B. Likewise, the rover on the right can figure out the message A.

In this scheme, the satellite only needs to transmit one message instead of two messages, that is, a 50% savings in transmission power. This is crucial because power is a scarce resource on a satellite.

Therefore, Network Coding can do much better than store-and-forward.

A similar application of Network Coding can also be used for wireless communications. In fact, a scheme proposed at CUHK has been prototyped on a Wi-Fi network by MIT, showing very promising results.Besides communications, Network Coding also applies to data storage.

Suppose we have 4 data hard disks. To back up the data, the conventional solution requires 4 extra hard disks. A more frugal solution uses only 1 extra disk that contains a mixture of the data on the original disks.

If any of the disk breaks down, then the lost data can be recovered the other 4 disks.

This simple scheme can be viewed as a special case of Network Coding. Network Coding theory is only beginning to tell us what are the fundamental tradeoffs in data storage. Here is yet another field where network coding has left its mark. Microsoft has used Network Coding to develop its own version of BitTorrent to share files on the Internet.

A similar application for streaming video has been deployed by UUSee.com, a major content distributor in Mainland China.

There are many more fields where Network Coding can be applied. Examples are: wireless communication networks, sensor networks such as for defense applications, telephone calls over the Internet, and TV broadcast over the Internet.

And performance improvements by Network Coding could mean faster downloading of files, better mobile phone reception, better quality in streaming video, and more privacy in communication.

To summarize, in many many problems concerning data storage, processing, and communication over networks, application of Network Coding can improve the performance, sometimes drastically. Very much like turning an old car into a brand new Ferrari.*With all these possible improvements Network Coding can make on communication systems, it is really a potential revolution as said in this special report on Network Coding in the Jun 2007 issue of Scientific American. Indeed, this revolution is already happening at an increasing speed.*Network Coding is a new paradigm because it fundamentally changes the way a communication network can operate. With Network Coding, we can build a new information superhighway system.*It is therefore not surprising that Network Coding has made tremendous impact on many fields in information science. These fields are shown on the right hand side.

The impact of Network Coding goes as far as mathematics, physics and biology.

In this AOE proposal, we will focus on fields that are related to information science.The number of research papers on Network Coding is growing extremely fast.The citation of our seminal paper has exceeded 1,350, with more than 500 papers written within the past 10 months.There are 4 books on Network Coding. One of them is my textbook on Information Theory and Network Coding that has been adopted by over 60 universities, including Stanford, UC Berkeley, and UC San Diego.And there are 7 special issues on Network Coding by major journals.There have been 6 best paper awards for Network Coding papers; 2 of them were won by this AOE team.There are now 2 annual international conferences on Network Coding.And Network Coding courses are taught at universities at universities around the world, such as MIT, Caltech, and UC Berkeley. Furthermore, many professors at leading universities are building their careers on Network Coding.

In addition to academia, Network Coding is being heavily researched by notable groups in industry, such as Bell Labs, Microsoft Research, and also DARPA, a defense agency of the US.

*Network coding has appeared in Wikipedia since 2006. There have been special reports on Network Coding in Scientific American and New Scientist.Our team has 23 members, with 5 IEEE Fellows, 13 best paper awards, 69 US patents with 1 patent on Network Coding.Here are our members in Hong Kong.And our members in Mainland China, Taiwan, and the US.Our industrial members are from Microsoft Research, Cisco, and Qualcomm. Cisco owns 70% of the computer router market worldwide, and Qualcomm developed the 3G mobile system.CUHK is the founding place of Network Coding.As I said earlier, our first paper is the most cited in the field. Another paper of ours is the second most cited paper, and this paper won the 2005 IEEE Information Theory Society Paper Award.It was the first time in more than 30 years that this award was won by Asian researchers.Ever since the inception of Network Coding, we have been obtaining seminal results that start most of the important branches in the field.Specifically, we are in the process of integrating the classical fields of channel coding theory, computer networks, switching theory, and cryptography into a modern theory of communication under the framework of Network Coding.We are also in the process of further developing the theory of Network Coding at the signal level in order to exploit the potential of Network Coding in wireless communications.Why do we need an AOE in Hong Kong?

The field of Network Coding is already very largePeople are forming conglomerates, and huge military projects are funded in the US and Europe.With just a few core people, we have been able to lay the foundations of Network Coding. However, in order to expand our leadership, we need the human resources to work on the large number of follow-up research problems and emerging applications.

An Institute of Network Coding is the answer.This AOE team can do the job, because we have very unique strengths:

On the academic side, we have the strongest theory group in the field and also a wide coverage of expertise in communications research. On the industrial side, we team up with 3 technology giants in IT. Therefore we are the only team that can support a complete knowledge supply chain from theory to practice.And what are the advantages of Hong Kong?

Being a cosmopolitan city, numerous researchers visit Hong Kong and give seminars all the time. This is a very big advantage for information exchange. There is a strong telecommunication manufacturing industry in Shenzhen, with major companies like Huawei and ZTE. We in Hong Kong are in a great position to help China as a whole to engage in this fast growing field and to capture the opportunity. Now let us talk about the goals of our AOE.To sustain our global leadership, we will keep abreast of all existing areas and steer the field into new directions.To make Hong Kong the international hub, we will invite eminent scholars to visit the Institute and hold international conferences.The Institute of Network Coding will train PhDs and postdocs who are employable by top universities and research labs. The PhD programs in Hong Kong have yet to be internationally recognized, and this AOE will contribute in this regard. To boost regional interest in Network Coding, we will establish a consulting arm to extend outreach and impact of the Institute.To attract major technology companies to set up research labs in Hong Kong, we need a showcase of Hong Kongs potential in hi-tech. The Institute of Network Coding will serve this purpose. As a matter of fact, I have been working with the Hong Kong Science Park to attract IBM and Cisco to Hong Kong.In preparation of this AOE, we formed the Network Coding Research Center at CUHK in 2006. Among the ground work we have done, we held the international Network Coding Workshop in Hong Kong in 2008.

Riding on the wave of Network Coding, many excellent PhDs are being produced around the world. Establishing the Institute of Network Coding now, we will be able to hire many of them because the current job market is poor. However, the window of good timing to capitalize on all of these, is quite narrow.Deliverables. According to our goals, this AOE has the following deliverables.

First, we will establish the Institute of Network Coding as the research infrastructure. For postdoc hiring, although we will be competing with very famous universities, I am confident that we will be able to attract many top talents if we act immediately.

Secondly, we plan to hold a number of international conferences, including the flagship Information Theory Symposium in Hong Kong in 2013 or 2015.

Continuing with the deliverables Under the umbrella of the Institute, we will hold seminar series on Network Coding. We will transfer Network Coding knowledge to industry. We will build system prototypes that can be subsequently picked up by the market, and along the way generate IPRs and incubate spin-off companies.

To support our deliverables, we plan to have application projects with potentially significant impact in the next 5 to 10 years.

As examples, on the upper left we have a project on the next generation Internet Protocol-TV system with Network Coding. On the top, we have a project with Cisco to re-engineer the Internet backbone with Network Coding.And on the upper right, we have a project with Qualcomm to study the application of Network Coding in a wireless network.

*To support our research projects, we will build a number of test-beds for evaluating Network Coding protocols. These test-beds should be up and running within the first 2 years.

We are also considering a campus-wide wireless test-bed on the CUHK campus.*Our budget request, shared by 15 local Co-Is over 8 years, is approximately 100 Million Hong Kong Dollars. 77% of the budget is for salary, and about 7% for equipment and prototyping. We also have a budget item on teaching buyout for the PIs.*We plan to hire 4 Research Assistant Professors, and there will be 4 positions for Visiting Research Fellows. There will be 10 concurrent Postdocs supervised by our 15 local investigators. And we will hire at least 2 full-time engineers for building systems and prototypes.*For the governance structure, there are 4 bodies in the AOE as shown here, with the following functions.*The Advisory Board consists of the Vice-Chancellor and Bursar of CUHK and the Project Coordinator -- myself. This board deals with fiscal matters.

The Scientific Advisory Board has 3 to 5 prominent scholars in related fields, one of them being Prof. Andrew Yao, winner of the Turing Award in Computer Science. This board deals with science and engineering matters and long-term directions, and is responsible for promoting visibility of the Institute. **The Executive Board of Governors consists of the 2 P-Is and 3 senior academic or industrial members, two of them being Prof. Vincent Chan, Chair Professor of MIT, and Dr. Nim Cheung, CEO of the Hong Kong Advanced Science and Technology Research Institute (ASTRI).

This board has ultimate control of the AOE. It sets the directions of the AOE, approves the overall budget, sets up ad hoc committees, and endorses the annual & progress reports.**The Research Committee consists of all P-Is and Co-Is. It carries out directions from the Executive Board of Governors, deals with routine matters, proposes and consolidates projects, and organizing seminars series on Network Coding.**The future of this AOE beyond its supporting period depends on different scenarios.

On the right is the unlikely scenario that at the end of this AOE, all ideas in Network Coding have been developed and productized. Then the AOEs mission is positively accomplished.

Under the scenario on the left, if the AOE is going strong, then it should be fairly self-sustainable through funding from the private sector and the government, such as the ITF.

The last scenario is such that there are new major breakthroughs along the way, which is what we are hoping for.In this case, the AOE should evolve into an entity of a larger scope through different sources of funding, including private donations.This would be similar to the Institute of Mathematical Sciences at CUHK, headed by Fields Medalist Prof. Shing-Tung Yau.*In conclusion,

We need an AOE to elevate our global leadership in Network Coding to a higher level.

We hope to bring substantial benefits to Hong Kong by providing the base for research, education, industry, and most importantly economy by fostering a capacity for knowledge pipelining.

Together, we will make this Hong Kong success story even bigger by positioning ourselves at the forefront of this next revolution in information science and technology.

Finally, I would say that Hong Kong has the legitimate claim to be the Mecca of Network Coding.*We are very excited about this opportunity.Thank you.