NeXtworking03 June 2325,2003, Chania, Crete, Greece

1
Summary IIQuestions we Hope to Answer
(non-wireless)

• Ioannis Stavrakakis, Univ. Of Athens
• Ibrahim Matta, Boston Univ.
• Michael Smirnov, Fraunhofer FOKUS

2
CS
SUC
3
(No Transcript)
4
NewArch

• Cross Layer is NOT ONLY wireless/sensor issue
• Patterns templates for arch. Instantiation,
  or/and
• programmable interfaces between layers
• Architecture on demand ? architectural layer
  becomes a design variable
• Non-layerd architecture, e.g. RBA ICIR
• If cross layer becomes a reality
• Assuming cooperative behaviour ... do we know how
  to detect behaviour violations?
• Auditing based on high-level invariants

5
Trends
Traditional control
Real time
Network Control Trend
Technology needs to manage itself Irving
Wladawsky-Berger, IBM
Self controlled
reactivity
Network Mngmnt Trend
system-width
Self-managed
Traditional management
6
Complexity Trends
System Syze Large Scale
State Mngmnt granularity
7
Need Feedback

• NeXtworking 2003 guidelines
• Concentrate on fundamentals
• Cross-cultural (interdisciplinary) approach
• Cross-awareness and self-everything,
• Expected result understanding of needs
• NeXtworking 04 (if annual)
• in USA
• Keep only one topic but attack it from many
  viewpoints
• Topic -sensor networks
• wireless, mobile, large-scale, ...

NetworkingSyncretism
8
Report Issues (focus on Funding Bodies)

• Implementation strategies
• Topics questions
• Filter out Technology Noise IS
• Which neighbouring disciplines may help
• Applied statistics, Science of Design TZ,
  Model driven engineering, Systems research,
  component-based systems, Coordination theory, ...
  , Distributed , autonomic, selfware, ...,
  Internet economy
• Performance Modalities
• Exchanges
• May extend e.g. E-NEXTs Virtual Laboratory to
  host exchanges in EU
• PhD students selection
• Testbeds and data repositories
• JRP
• Do EU and US officials need a framework agreement?

9
Recommendations

• Need measurable objectives
• E.g record each participants weight before and
  after the event
• Make PowerPoint timer programmable
• ...

10
Original Questions
11
Traffic Modelling and Control

• Is standard queueing theory enough?
• Is first-course control theory enough?
• How can we understand time-scale interactions and
  design for them?
• Are time-varying averages enough, or do we need
  to capture higher moments?
• What happens to effective banwidth and good old
  Erlang?
• Why we are insisting on modelling TCP and
  RED?
• Why did it take some time to be more careful?
• Self-similarity vs. pseudo-ss, limit cycle vs
  instability, etc.

12
Peer-to-Peer and CDNs

• Whats new about P2P? Is it a buzzword?
• How is different from/similar to distributed
  computing?
• Is lookup/indexing the hard P2P problem?
• Is it solved with DHTs?
• Overlays will it ever end?
• Could we afford ignoring underlays?
• Do we expect peering of overlays to
  emerge/succeed?
• Is it a contradiction?
• CDNs is the D for Delivery or Distribution?
• is it only replication?

13
Network Tomography

• How to map application needs to the type and
  spatio-temporal granularity and precision of
  tomography?
• Are end-to-end measurements enough?
• What are the fundamental limits?
• What if the network does not support multicast?
• How can the results be validated?
• Do we need to or can we infer higher-order
  dynamics?

14
Network Measurements

• Data collection versus characterisation
• Conflict of interest?
• Need for trustworthdy datasets complete and
  representive, with context
• How can we promote sharing data?
• Characterisation versus understanding
• Characterisation should be motivated
• curve fitting is not enough the science is in
  the understanding of the characteristics
• How invariant are invariants?
• What about on-line measurements?

15
Programming the (Inter)Network

• Do we now how to do it predictably?
• Safety complexity comes from composition
• Tussles trust, accounting, QoS, etc.
• Do we need help from PL/OS communities?
• Do we have a common programming language?
• Compiler techniques, typing, etc.
• Intel-independent network processor architecture

16
Large-scale Architectures

• Why dont we just uphold the E2E principle?
• How can we succeed without a business model?
• Do we need help from economists?
• Do we really need a single integrated network?
• Will we ever achieve stable, adaptive routeing
• Should we embrace autonomic computing/networking
  
• Sekf-healing, self-organizing, self-aware etc.
• De we have the tool(s) to evaluate such large
  systems?