The Challenge of Challenged Networks S Keshav University

The Challenge of Challenged Networks S. Keshav University of Waterloo Dharamshala, September 8, 2010

Waterloo? Where is that?

Home of: RIM/Blackberry Watcom/Sybase Maple Open. Text Manu. Life

The Challenge of Challenged Networks S. Keshav University of Waterloo Dharamshala, September 8, 2010

A Fundamental Problem n Technology operates in a social, cultural and economic context n Real-world success depends on both the technology and the context n Nearly impossible to determine context in advance – and it may change

Challenged networks n Challenging conditions create unique requirements and also place strong constraints including: – – – Environmental Infrastructural Social Cultural Legal Economic

How to proceed? n Iterative process of design, deployment, and refinement

Context: requirements, constraints, and alternatives Design Redefine requirements and constraints Implement Test Monitor Deploy

Context: the Soft. Bridge stack People UI Synch Media Device Network Power Tucker and Blake, University of the Western Cape, South Africa

Design n Design every layer of the Soft. Bridge stack, not just the network! n DTN offers alternatives but is not a panacea n Don’t underestimate cellular networks – economies of scale – femtocells

Implement n Challenged networks limit options – avoid lock-in n Don’t build on unproven technology – ‘error 22’ (? ) n Build in monitoring

Test n Don’t skimp on testing, painful though it may be n Build test harnesses n Always investigate unexpected results – What you cannot explain is at the heart of new knowledge

Deploy n Always start with a pilot n Feet on the street n Start small

Monitor n Monitor everything n Visualize data graphically n Don’t skimp on analysis n Insist on formal reporting

Refine n Use monitoring data to refine assumptions, constraints, and requirements n Be prepared to throw away the first and second systems – ruthlessness is essential

The rule of three n The first system is too naïve n The second system is too ambitious n The third system is usable

Another rule of three n The first system is too simple n The second system is too complex n The third system is simple in just the right ways Strive to achieve the second simplicity!

Summary so far n Contextualize n Iterate n Simplify

Case study: Kiosk. Net

Kiosk. Net (2005 -2008) Trade delay for cost

People UI Synch Media Device Network Power

Experiences n Single-board Linux-based computers are not mass-market – hard to debug and maintain by locals n Vehicular environment is harsh – failure – theft n Difficult to get agreement from transportation providers

VLink (2009) n Addresses problems with Kiosk. Net – – – n Reuses existing Windows and Linux desktops Software-only solution No computer in vehicle No need for buy in from transportation authorities and MUCH cheaper! Leverages USB memory sticks (Key. Link) and SMS (SMSLink)

Applications n Email n VSync – keeps two directories in sync

Outcome? n Probably a failure n NGOs do not want unsupported software n SMS/voice communication on mobiles appears adequate – 3 G is coming

Case study: TIER/Aravind People UI Synch Media Device Network Power

Conclusions n n n Designing any successful network is difficult; challenging environments exacerbate this Contextualize using the Soft. Bridge stack Iterate to build at least three solutions Simplify the systematically and ruthlessly Success is not assured

Thank you! Grad students : S. Liang, A. Seth, N. Ahmed, M. Ghaderi, S. Guo, M. H. Falaki, S. Ur Rahman, E. A. Oliver, U. Ismail, S. Rahbar n Graduate interns: R. Luk, Z. Koradia n Staff Programmer: D. Kroeker, M. Derakhshani, A. Ganjali n Undergrads : M. Zaharia, P. Darragh, N. Arora, Y. Yin, G. Salmon, G. Wang, M. Liang, M. Thomas, A. Agarwal, Y. Xu, Y. Hu, S. Dube, R. Sethi, C. Ho, C. Tan, A. Leong n Affiliated Faculty: T. Brecht (UW), U. Hengartner (UW), S. Prasad (IIT Delhi), H. Saran (IIT Delhi) n Staff support: G. Chopiak n