IP Switching for Scalable IP Services Hassan M. Ahmed Ross Callon Andrew G. Malis Hohn Moy

IP Switching for Scalable IP ServicesHassan M. Ahmed

Ross CallonAndrew G. Malis

Hohn Moy

Presented by Gao, YunShih, Pei-ShinWei, ShuGuang

OUTLINE Background Review & Motivation The Overlay Model: Classical IP

over ATM IP Switching IP Navigator Conclusion

Background Review & Motivation Hop-by hop routing

Simplicity Hierarchical Routing, easy to scaling Difficult to implement Traffic

Engineering (Bandwidth Management) & QoS

Background Review & Motivation (Continued) Switched Core

Achieve a form of Traffic Engineering VC’s allows Explicit Routing to be used

efficiently Isolate the internal routing from

changes of the Internet’s routing algorithms

Integration of Datagram and Circuit Technologies

The Overlay Model: Classical IP over ATM IP vs. ATM

Connectionless (IP) vs. connection oriented (ATM)

Packets (IP) vs. cells (ATM) Broadcast LAN’s (IP) vs. point-to-point

connections (ATM)

The Overlay Model: Classical IP over ATM (Continued) ATM Address Resolution Protocol

(ATMARP) Logical IP subnet (LIS) Independent Routing Protocols

Host A

Host B

Router 1

Router 2

LIS 1

LIS 2 LIS 3

Host C

Host A

Host B

Router 1

Router 2

LIS 1

LIS 2 LIS 3

Host C

Multiple IP LIS’s on one ATM network Connection between Hosts A and B

ATMSVC

The Overlay Model: Classical IP over ATM (Continued) Logical IP subnet (LIS)

IP stations (hosts & routers) in the same LIS communicate directly via ATM SVC’s or PVC’s

IP stations in different LIS’s must intercommunicate via a router

Next Hop Resolution Protocol (NHRP) Query / Response Model

Host A

Host B

Router 1

Router 2

LIS 1

LIS 2 LIS 3

Host C

Host A

Host B

Router 1

Router 2

LIS 1LIS 2 LIS 3

Host CATMSVC

Direct Connection between Hosts A and C (NHRP)Connection between Hosts A and C

ATMSVC

ATMSVC

ATMSVC

The Overlay Model: Classical IP over ATM (Continued)

Scaling Problem Total number of logical links that are

advertised between the n ATM-attached routers equals

2)1(* nn

IP Switching Eliminating scaling problems by

running the IP routing protocol on switches as well as routers

IP Navigator1. A particular IP switching

implementation2. Developed by Cascade

Communication Corporation

IP Navigator (Continued) Makes a “cloud” of Cascade

switches, frame relay, or ATM Appears externally to be a

collection of IP routers

A

FE

D

C

B

I

G

H

IP Navigator (Continued) Two routing instances are running

inside the “cloud” Uses standard IP routing (OSPF)

within the core to exchange routing information

A VC routing protocol is running between switches, allowing them to build up point-to-point and point-to-multipoint VC’s

IP Navigator (Continued) Each router pre-establishes a VC to

each potential egress (i.e. to every other router in the area) Build point-to-multipoint (PMT) tree

rooted at each egress Traffic travels in reversed direction

VC’s used by IP Navigator are set up in response to routing packets and are automatically re-established as necessary

IP Navigator (Continued) Multicast Similar to unicast Standard IP multicast protocol are

spoken at the edge of the cloud Multicast information is

redistributed throughout the cloud using OSPF

A

FE

D

C

B

I

G

H

Example of multipoint-to-point tree (MPT)

QoS and Traffic Engineering VC routing is based on dynamic

routing algorithm Explicit routing allows

1. Crankback and retry2. Optimization of the combined path

for multiple VC’s

QoS and Traffic Engineering (Continued) IP Navigator allows QoS support to

be based on a range of coarse through fine granularity Traditional “best efforts” IP service Separates IP traffic into a small

number of classes and open separate MPT’s for each class

First Class Economy Class

Conclusions IP Navigator integrates the

transport of connectionless IP traffic over connection-oriented switched data networks

Better scaling properties and inherent simplicity from IP

Higher performance of forwarding packets (VC’s)