Impressions (looking over the fence)

1
Impressions(looking over the fence)

• Jim Kurose
• Department of Computer Science
• University of Massachusetts
• Amherst MA 01003

2
Impressions lots of interesting talks

• application use
• improved PHY, MACs, routing
• diversity radios, antenna, paths
• deployment experience
• applicability of simulation (?)
• security
• management

Interesting talks by interesting people no
surprises!
3
Surprises what are the apps?

• Multiple application uses (from Victor Bs talk)
• within neighborhood (backup, video, games, VoIP)
• multihop to wired network
• networking in infrastructure-poor countries
• from todays panel .
• homeland security, military apps, sensing,
  multihop connectivity to wired network from
  industry panelists

4
Surprises what are the apps?

• Multiple application uses (from Victor Bs talk)
• within neighborhood (backup, video, games, VoIP)
• multihop to wired network
• networking in infrastructure-poor countries
• but
• does one of these dominate as a killer app?
• do requirements change from one to another?
• deployment scenarios, application-use driving
  research agenda different from 30 yrs of ad hoc
  net research?
• smaller (non-global scale)
• zero conf

P Kumar we dont know killer app right
architecture is why Internet was a success
5
Surprises whats the service model?

• is there (implicit) agreement
• best effort, or some form of QoS which?
• lessons from 15 years of internet QoS, but
  different setting here?
• to what extent do protocols, architecture support
  service model, applications?
• focus mostly on connectivity, capacity
• 802.11 access, routing good enough?
• solving the general, or the specific?

6
The ities

• evolvability if important apps unknown, how does
  architecture support evolvability?
• lessons from Internet?
• manageability if zero-management usability, how
  does architecture support this?
• trading performance for manageability, (e.g.,
  dedicating resources to improve manageability
  separating control and data paths)
• how to quantify/characterize this tradeoff?

Q How are these goals reflected in protocols,
architecture?
Opportunity address these issues from scratch!