Panel Contribution: The Future of Transport

1
Panel ContributionThe Future of
TransportEvolution or Revolution

• Steve Deering
• deering_at_cisco.com

2
Above My Layer!

• transport protocol development is crucial to
  restore the thin waist of the hourglass
• an architecturally clean alternative to various
  transport helper hacks inside the network
• datalink protocol development also important
• multicast transport development needed to remove
  obstacle to IP multicast deployment
• will become even more essential if end-to-end
  encryption (IPsec) becomes common

3
Better IP Layer Supportfor Transport Protocols?

• disincentives for bad (non-cooperative) behavior?
• e.g. penalty boxes
• definition/characterization of bad behavior
• better fairness?
• e.g., stochastic fair queueing
• better congestion feedback?
• e.g., ECN

4
Challenging Link/Path Properties(Vern Paxsons
list)

• long delayaffects timer estimation, congestion
  adaptation
• high bandwidth-delay productdifficult to fully
  utilize and still respond to congestion
• variable bandwidthdifficult to track throughput
  target
• varying delaymay affect accuracy of RTO
  estimation

5
Challenging Link/Path Properties(cont.)

• links with link-layer flow controlmay interact
  adversly with end-to-end control
• high error ratesperformance effects when
  interpreted as congestion
• inconsistent error ratescan defeat attempts to
  tell loss from congestion
• expensive connection set-up or retentionhostile
  to soft-state protocols

6
Challenging Link/Path Properties(cont.)

• very low bandwidthcreates pressure for blurring
  layers and custom demuxing
• unidirectionalbreaks most current IP routing
  (not really a transport issue)
• non-transitive reachabilitybreaks subnet
  assumptions (not really a transport issue)
• large clouds(not really a transport issue)

7
Challenging Link/Path Properties(cont.)

• paths that reorder packetsdefeats common
  optimizations, hueristics
• multipathingmultiple path properties to estimate
• intermittent outagesmay excessively stretch RTT
  estimates
• very small MTUmay create black holes if MTU
  discovery doesnt work

8
Challenging Link/Path Properties(cont.)

• Question is, which of these can/should be
• handled at transport layer?
• handled at datalink layer?
• ignored?

9
One Pragmatic Suggestion

• use ALF (or at least above-UDP) approaches to
  enable easier deployment