Ten Gigabit Ethernet Research Questions

Slides: Ten Gigabit Ethernet
Research Questions

[Start Here (html)] [Slide Sorter (html)]


Page 1: Ten Gigabit Ethernet

Page 2: Outline

Page 3: The Web100 project

Page 4: The Wizard Gap

Page 5: The Wizard Gap

Page 6: TCP "tuning" requires experts

Page 7: TCP tuning is painful debugging

Page 8: Web100: Just ask TCP

Page 9: Why is this important?

Page 10: Further changes

Page 11: Loose ends for further research

Page 12: Protocol Interactions

Page 13: Framing the problem

Page 14: Two Primary TCP Control Systems

Page 15: Congestion Control

Page 16: TCP Self Clock

Page 17: Tcptrace

Page 18: A Tcptrace

Page 19: Capture Effect

Page 20: ATM Policer Failure

Page 21: TCP Correct Operations Summary

Page 22: MTU Issues

Page 23: Quote

Page 24: Moore's Law

Page 25: Packet Times

Page 26: Costs #1: Cycles to Process Packets

Page 27: Costs #2: PSC's TCS

Page 28: Why not Ethernet in the TCS?

Page 29: Machine Area Nets

Page 30: Need to Consider All Costs

Page 31: The ATM rat hole

Page 32: Proposed retro History

Page 33: How do we get from here to there?

Page 34: A Word About Cost

Page 35: A Prediction About Benefits

Page 36: Wrap up

Page 37: MTU Compounds Protocol Problems

Page 38: Wizard Gap Revisited

Page 39: Wizard Gap Revisited

Page 40: Which is more important?

Page 41:

Page 42:

Page 43: Diagnosing path bottlenecks

Page 44: Web100 Project Overview

Page 45: Current Web100 status

Page 46:

Generated by MagicPoint