CITM600 Chp 7

gascitytankΔίκτυα και Επικοινωνίες

28 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

89 εμφανίσεις

Top
-
Down Network Design


Chapter Seven:

Selecting Switching and Routing Protocols




Ted Rogers School of Information Technology Management

Ryerson University

Objectives


To learn about different routing protocols



Guide you in selecting the right routing
protocol for the network design



Develop a systematic process for selecting
the right solution



Comparison between routing and switching in
network applications

©

R.
Hudyma

Switching and Routing Choices

Switching:


Layer 2 transparent bridging (switching),


VLAN technologies.


Multilayer switching,


Spanning Tree Protocol enhancements,

Routing:


Static or dynamic,


Distance
-
vector and link
-
state protocols,


Interior and exterior, etc…

©

R.
Hudyma

Selection Criteria for Switching and Routing Protocols


To help to select the right routing
protocol, we will consider the following:


Network traffic characteristics.


Bandwidth, memory, and CPU usage.


The number of peers supported.


The capability to adapt to changes
quickly.


Support for authentication.

©

R.
Hudyma

Making Decisions


Four major factors in decision making:


Goals must be established.


Many options should be explored.


The consequences of the decision should be
investigated.


Contingency plans should be made.



To help facilitate this process, a decision table
can be used (Table 7.1, Next slide)

©

R.
Hudyma

Example Decision Table

©

R.
Hudyma

Troubleshoot the Decision


Ask Yourself the following after making a
decision to help solidify your decision:



If we use the selected option, what could go
wrong?


Has this been tried before? What were the
problems or challenges?


How will the customer react to the decision?


What is the contingency if the customer does
not approve, (Backup Plan)


©

R.
Hudyma

Transparent Switching


Ethernet Switches have the same characteristics as Bridges


Take Advantage of Fast Integrated Circuits to reduce latency
associated with Bridging.


Higher port density and lower cost than bridges due to Integration


Usually implement Cut Through switching:


Switch looks only at the Destination Address

(First Field in the LAN Frame)


Looks for the destination port in the MAC address table


Start sending the frame to the destination port immediately without
waiting for the entire frame


Disadvantages of Cut
-
Through Switching, bad frames get forwarded
since the get queued as soon as they are received without waiting for
CRC Checks


Adaptive Cut Through Switches moves to store and forward when a
certain error threshold is reached.


©

R.
Hudyma

Transparent Bridging (Switching) Tasks


Forward frames transparently


Operates at Layers 1 and 2 of the OSI Layer


Creates Bandwidth Domains not Broadcast Domains


Learn which port to use for each MAC address using:


Switching table


MAC Address Table


CAM table


Flood frames when the destination unicast address
hasn’t been learned yet.


Filter frames from going out ports that don’t include the
destination address.


Flood broadcasts and multicasts.


Bridges are store and forward devices

©

R.
Hudyma

Switching Table on a Bridge or Switch

MAC Address

Port

1

2

3

08
-
00
-
07
-
06
-
41
-
B9

00
-
00
-
0C
-
60
-
7C
-
01

00
-
80
-
24
-
07
-
8C
-
02

©

R.
Hudyma

Protocols for Transporting VLAN Information


Inter
-
Switch Link (ISL):


Encapsulation protocol for VLAN information


Cisco proprietary, mostly used to interoperate with older Cisco
Hardware


Newer Cisco hardware supports both and sometimes only 802.1q

IEEE 802.1q:



Tagging protocol, Tag is added to the front of the packet.


IEEE standard.


Supported by Most Manufacturers


Changes the Frame so FCS must be re
-
calculated

VLAN Trunk Protocol (VTP):


Client Server VLAN management protocol


Pushes VLAN configuration from Server to all Client Switches


Can be problematic in large environments.

Dynamic Trunk Protocol (DTP):


Allows dynamic negotiation of 802.1q trunk parameters.


Can be difficult to configure due to the number of options available.


Better to configure the trunk parameters manually and consistently.

©

R.
Hudyma

Cisco Multilayer Switching


High speed routers are used with Layer 2 network
and are known as Layer 3 switches


Cisco’s implementation contains the following:


Route processor or router


Switching engine


The Multilayer Switching Protocol (MLSP).


Establishes the concept of flows


First packet is routed to the right port on the switch


All subsequent packets for that specific flow are
switched at layer 2 to the destination port until the
flow changes

©

R.
Hudyma

Spanning Tree Protocol Enhancements


802.1w Rapid Spanning Tree


802.1s grouping of VLANs into a single Spanning Tree instance


PortFast (Switch Edge Port)


Normal STP startup procedure can take 30
-

50 seconds.


Blocking, Listening, Learning


Good for network devices, bad for workstations.


DHCP Timeout, Domain Controller timeout


PortFast allows you disables the STP algorithm for ports that are known to
be workstations, The port immediately goes into forwarding mode


UplinkFast and Backbone Fast.


UplinkFast improves STP recovery from 30
-
50 seconds to 1 second


BackBone Fast can save 20 seconds when recovering from a link failure


Unidirectional link detection and Loop Guard


Prevents connection that are only broken in one direction.


BPDU’s could flow in only on direction creating a loop.

©

R.
Hudyma

Redundant Uplinks

Access
Layer

Distribution
Layer

Core

Layer

Switch A

Switch B

Switch C

Primary
Uplink

Secondary
Uplink

X

X

X

= blocked by STP

If a link fails, how long will STP take to recover?

Use UplinkFast to speed convergence

©

R.
Hudyma

Selecting Routing Protocols


They all have the same general goal:


To share network reach
-
ability information among
routers.


They differ in many ways:


Interior versus exterior,


Metrics supported,


Dynamic versus static and default,


Distance
-
vector versus link
-
sate,


Classful versus classless,


Scalability.

©

R.
Hudyma

Distance
-
Vector Routing


Router maintains a routing table that lists known networks, direction
(vector) to each network, and the distance to each network.


Router periodically (every 30 seconds, for example) transmits the
routing table via a broadcast packet that reaches all other routers on
the local segments.


Router updates the routing table, if necessary, based on received
broadcasts and distributes the new information to its neighbor at the
next 30 second interval


Usually implements Split
-
Horizon, do not advertise routes that it learned
from a port out that port, only routes that were learned from other ports.


Also implement hold down timer, Delays accepting new information into
the routing table to make sure that the routes are not based on stale
information


Implements a hop count limit to stop packets from looping forever


Eg. RIPv1 and v2, IGRP/EIGRP, Novell IPX RIP, AppleTalk RTMP)

©

R.
Hudyma

Distance
-
Vector Routing Tables

Router A

Router B

172.16.0.0

192.168.2.0

Network

Distance

Send To


172.16.0.0


0


Port 1

192.168.2.0


1


Router B


Network

Distance

Send To


192.168.2.0


0


Port 1
172.16.0.0


1


Router A


Router A’s Routing Table

Router B’s Routing Table

©

R.
Hudyma

Link
-
State Routing


Do not exchange routing tables, only information about links
connected to the router.


Routers Learn About links in the network from their neighbor
routers.


Routers send updates only when there’s a change.


Router that detects change creates a link
-
state advertisement
(LSA) and sends it to neighbors.


Neighbors propagate the change to their neighbors.


Routers update their topological database if necessary.


Uses a shortest path first algorithm like Dijkstra’s algorithm.


More CPU and memory intensive than distance vector but
bandwidth friendly.


Eg. OSPF, IS
-
IS and Novell’s NLSP.


©

R.
Hudyma

Distance
-
Vector vs. Link
-
State


Distance
-
vector algorithms keep a list of networks, with
next hop and distance (metric) information.


Most of the implementations sends the entire routing
table every update (IGRP/EIGRP do send incremental
updates when changes )


Link
-
state algorithms keep a database of routers and
links between them:


Link
-
state algorithms think of the internetwork as a
graph instead of a list,


When changes occur, link
-
state algorithms apply
Dijkstra’s shortest
-
path algorithm

to find the shortest
path between any two nodes.

©

R.
Hudyma

Choosing Between Distance
-
Vector and Link
-
State

Choose Distance
-
Vector


Simple, flat topology


No Hierarchy


Simple Hub
-
and
-
spoke
topology


Network Administrator
do not have the skills to
troubleshoot a Link
State Protocol.


Convergence time not a
big concern


No redundancy

Choose Link
-
State


Large Network


Hierarchical topology


Network Administrator
skill levels is adequate


Redundancy and Fast
convergence is critical

©

R.
Hudyma

Dynamic IP routing protocols

Distance
-
Vector


Routing Information
Protocol (RIP) Version
1 and 2


Interior Gateway
Routing Protocol
(IGRP)


Enhanced IGRP


Border Gateway
Protocol (BGP)

Link
-
State


Open Shortest Path
First (OSPF)



Intermediate System
-
to
-
Intermediate System
(IS
-
IS)

©

R.
Hudyma

Routing Protocol Metrics

Metric: the determining factor used by a routing algorithm
to decide which route to a network is better than another.


Examples of metrics:


Bandwidth
-

capacity


Delay
-

time


Load
-

amount of network traffic


Reliability
-

error rate


Hop count
-

number of links/routers that a packet
must travel through before reaching the destination
network


Cost
-

arbitrary value defined by the protocol or
administrator

©

R.
Hudyma

Interior Versus Exterior Routing Protocols


Interior routing protocols are used within an autonomous
system.


RIPv1, RIPv2, IGRP, EIGRP,


Exterior routing protocols are used between autonomous
systems.


BGP4


Autonomous system (two definitions that are often used):

“A set of routers that presents a common routing policy to the
internetwork”

“A network or set of networks that are under the administrative control
of a single entity”



©

R.
Hudyma

Classful vs Classless Routing Protocols


Classful protocol make assumptions about
the subnet mask based on traditional IP
Classes A, B, C.


RIPv1, IGRP are examples of classful protocols


Classless Routing protocols transmit the
prefix with the routing updates so can support
VLSM and CIDR.


RIPv2, EIGRP,OSPF, ISIS and BGP are examples
of classless routing protocols.

©

R.
Hudyma

Routing Algorithms

Static routing:


Calculated and Configured beforehand, offline.

Default routing:


“If I don’t recognize the destination, just send the
packet to Router X”, statically configured

Cisco’s On
-
Demand Routing:


Routing for stub networks.


Also used to engage dial backup.


Uses Cisco Discovery Protocol (CDP).

Dynamic routing protocol:


Distance
-
vector algorithms.


Link
-
state algorithms.

©

R.
Hudyma

Static Routing Example

RouterA(config) #
ip route 172.16.50.0 255.255.255.0 172.16.20.2

Send packets for subnet 50 to 172.16.20.2 (Router B)

e0

e0

e0

s0

s1

s0

s0

Router A

Router B

Router C

Host A

Host C

Host B

172.16.10.2

172.16.30.2

172.16.50.2

172.16.20.1

172.16.40.1

172.16.10.1

172.16.30.1

172.16.50.1

172.16.20.2

172.16.40.2

©

R.
Hudyma

Default Routing Example

RouterA(config) #
ip route 0.0.0.0 0.0.0.0 172.16.20.2

If it’s not local, send it to 172.16.20.2 (Router B)

e0

e0

e0

s0

s1

s0

s0

Router A

Router B

Router C

Host A

Host C

Host B

172.16.10.2

172.16.30.2

172.16.50.2

172.16.20.1

172.16.40.1

172.16.10.1

172.16.30.1

172.16.50.1

172.16.20.2

172.16.40.2

©

R.
Hudyma

Scalability Constraints for Routing Protocols

Questions to ask about Scalability of routing protocols


Are there limits on the metrics?


How quickly does the protocol converge in case of a problem?


How often are routing updates transmitted?


How much data is sent in a routing update? Whole table or
just updates?


How widely are routing updates propagated? Neighbor?
Whole network?


How much CPU or memory is used to process the updates?


Are static or default route supported?


Is route summarization supported?


Does the Routing protocol support Classless routing and
VLSM?



©

R.
Hudyma

Routing Protocol Convergence


Convergence is the time it takes for routers to
arrive at a consistent understanding of the
network topology after a change takes place.


Routers start the convergence process after it
detects a change in topology


Changes are detected in two ways:


Router can detect that an interface failed or lost
electrical signal


Missing periodic updates or hellos from the neighbour
router


©

R.
Hudyma

Routing Information Protocol (RIP)


First standard routing protocol developed for TCP/IP
environments:


RIP Version 1 is documented in RFC 1058 (1988).

RIP Version 1 Supports Classful routing Only.


RIP Version 2 is documented in RFC 2453 (1998).

RIP Version 2 Supports Classless routing.


Easy to configure and troubleshoot.


Broadcasts its routing table every 30 seconds; 25 routes
per packet.


Uses a single routing metric (hop count) to measure the
distance to a destination network; max hop count is 15.

©

R.
Hudyma

RIPv2 features


Includes the subnet mask with route updates:


Supports prefix routing

(classless routing, super
-
netting),


Supports variable
-
length subnet masking
(VLSM).


Includes simple authentication to foil crackers
sending routing updates.


Supports Route Tags to identify external routes
that were imported from another routing protocol.


©

R.
Hudyma

IGRP Solved Problems with RIP


IGRP was developed by Cisco to Overcome some of
the limitations of RIP.


IGRP supports 255 hops.


IGRP uses bandwidth, delay, reliability, load instead
of hop count


(By default just uses bandwidth and delay).


IGRP uses 90 seconds update intervals instead of
30 seconds.


Allows Load Balancing across equal cost paths.


Uses triggered updates to improve convergence
time.

©

R.
Hudyma

EIGRP (Enhanced IGRP)


Developed by Cisco to overcome the limitations of IGRP


Derived from and very similar to IGRP.


Adjusts to changes in internetwork very quickly.


Incremental updates contain only changes, not full routing table.


Updates are delivered reliably.


Router keeps track of neighbors’ routing tables and uses them as
feasible successor.


Uses DUAL (Diffusing Update Algorithm)


Same metric as IGRP, but more granularity

(32 bits instead of 24 bits)


Allows load balancing across equal cost paths


Supports triggered updates, Sending an update immediately in
response to a change to speed up convergence.


Carries prefix with routing updates so supports classless routing.

©

R.
Hudyma

Open Shortest Path First (OSPF)


Open standard, defined in RFC 2328 supported by many vendors.


Adjusts to changes and converges very quickly.


Authenticates protocol exchanges to meet security goals.


Supports discontinuous subnets and VLSM (Classless)


Uses Multicasts instead of Broadcast reducing CPU usage on LAN
hosts.


Supports very large Internetworks.


Has inherent hierarchical structure (Areas) that scales well to large
organized networks, reducing CPU and Memory requirements.


Does not use a lot of bandwidth (propagates only changes not the
entire routing table)


Uses hello packets every 10 seconds to ensure that neighbor
relationships are still there.


©

R.
Hudyma

OSPF Metrics

A single dimensionless value called
cost.
A network
administrator assigns an OSPF cost to each router
interface on the path to a network. The lower the cost,
the more likely the interface is to be used to forward
data traffic.


On a Cisco router, the cost of an interface defaults to
100,000,000 divided by the bandwidth for the interface.
For example, a 100
-
Mbps Ethernet interface has a cost
of 1
.


If interfaces larger than 100Mbps are used the default
costs should not be used, statically assigning cost or
changing the default auto cost reference value will
ensure that calculations are done properly.

©

R.
Hudyma

OSPF Areas Connected via Area Border Routers (ABRs)

Area 1

Area 3

Area 2

Area 0 (Backbone)

ABR

ABR

ABR

©

R.
Hudyma

OSPF Architectures


OSPF has natural hierarchy built into the
protocol


ABRs (Area Border Routers) Connect each area to the
Backbone Area (Area 0)


One OSPF Network is known as an AS

(Autonomous System).


ASBR’s (Autonomous System Boundary Router)
connect and Autonomous System to an external
network like the Internet.


Design OSPF with structure and hierarchy, design
Areas with IP Blocks that can be summarized


Supports stubby and not
-
so
-
stubby areas.

©

R.
Hudyma

IS
-
IS


Intermediate System
-
to
-
Intermediate System.


Link
-
state routing protocol, Similar to OSPF but
more flexible, efficient and scalable.


Designed by the ISO for the OSI protocols.


Integrated IS
-
IS handles IP also.


Hierachical protocol


Level 1 Routers route with an Area


Level 2 Routers route between Areas


Level 1
-
2 Routers are gateways between Intra Area and
inter area


Limited popularity and deployment


©

R.
Hudyma

Border Gateway Protocol (BGP4)


Developed by IETF to replace EGP (Exterior Gateway Protocol)


Allows routers in different autonomous systems to exchange routing
information:


Exterior routing protocol,


Used on the Internet among large ISPs and major
companies.


iBGP

(Internal BGP) is used to route between domains within
an AS


eBGP

(External BGP) is used to route to external networks like the
Internet


Supports route aggregation.


Main metric is the length of the list of autonomous system numbers,
but BGP also supports routing based on policies.


Main routing protocol of the Internet.


©

R.
Hudyma

Multiple Routing Protocols and Route Redistribution


You do not have to use the same protocol throughout the entire network


Routing protocols for the core should support


Redundant links


Load Sharing across equal cost paths


Fast Convergence


EIGRP, OSPF and ISIS


Routing protocols for the Distribution layer


RIPv2, EIGRP, OSPF and ISIS


Routing protocols for the Access layer


RIPv2, EIGRP, OSPF


Design Guidelines


Try to limit your choices to one IGP and one EGP


When there is requirements to support multiple protocols routes can be
redistributed between routing protocols.


Usually metric information is lost in re
-
distribution and decisions have to be
made to manually configure the treatment for the routes as they traverse into
the new network.


Administrative distance for each protocol further complicates the use of multiple
routing protocols See table 7
-
4


©

R.
Hudyma

Comparison Chart of Routing Protocols

©

R.
Hudyma

Summary


The selection of switching and routing protocols should be based
on an analysis of:


Goals,


Scalability and performance characteristics of the protocols.



Transparent bridging is used on modern switches:


But other choices involve enhancements to STP and protocols
for transporting VLAN information.



There are many types of routing protocols and many choices within
each type.



Chose the simplest routing protocol that will do the Job and be
scalable enough for your network

©

R.
Hudyma

Review Questions


What are some options for enhancing the Spanning
Tree Protocol?



What factors will help you decide whether distance
-
vector or link
-
state routing is best for your design
customer?



What factors will help you select a specific routing
protocol?



Why do static and default routing still play a role in many
modern network designs?

©

R.
Hudyma