CS551Watching the Waist of IP[Deering98a]Bill Chenghttp://merlot.usc.edu/cs551-f121 Computer Communications - CSCI 551 Copyright © William C. Cheng2The Protocol Hour Glass Computer Communications - CSCI 551 Copyright © William C. ChengIPTCP UDP ...SMTP HTTP RTP ...email WWW phone ...ethernet PPP ...CSMA async sonet ...copper fiber radio ...Discuss the role of the IP layer3Key Ideas Computer Communications - CSCI 551 Copyright © William C. Chengshould it grow or be replacedWhat does the IP layer need (according to deering)minimal functionalitycorollary: minimal functionality from link layermaximize interoperabilityWhy single:4Why A Single, Narrow Protocol? Computer Communications - CSCI 551 Copyright © William C. Chengminimize amount of work needed to support new protocolsminimize requirements from lower layersWhy narrow:end-to-end argument: don’t want to weigh down IP with abunch of things that are unnecessary by many of its usersconnectionless datagramSmall and simple5What Are The Key IP Properties? Computer Communications - CSCI 551 Copyright © William C. Chengmaximize connectivityGlobal addressingmuch harder to provide global addressability at higherlayerenable applications (e.g., peer-to-peer file sharing)but addressability make security harder => NAT boxes,firewallsQoS, multicast, ...More features are cooler6Why Add To Or Change IP? Computer Communications - CSCI 551 Copyright © William C. ChengMore features make more moneyReplacing it makes lots of moneyand could do new things if everyone buys ine.g., ATMHave short-term (?) problems that have to be solved (viawork-arounds)address space size: NATTCP "helpers", reliable multicast, packet-interceptingcaches, "content-based routing", active networkingSize of IP header: 40B (is this a problem?)interactions with ATM or other link-layers with short MTUsActive network: idea that end-users (or admins) should beable to reprogram the network7Other Questions/Observations? Computer Communications - CSCI 551 Copyright © William C. Chengcould be problem (high overhead) for telnetQoS and multicastwhy not deployed? increase assumptions lower layers, weneed to change all routers, not everyone needs services,difficult to transfer research into company, may be can dothese things above
View Full Document