Post on 14-Apr-2016
description
Network Routing Protocols andAdvanced Topics
Network Routing Protocols
Link & Frame Detail
Distributed Software Module
Node Software Applications
Exercise 3 - Client/Server Network
Packet Switching and Network Operation
Routing Protocols– Static and adaptive routing algorithms for packet switching– Metrics based on user-defined penalties, measured delays,
or composite metrics
Routing Classes– Routing protocols using shortest path metrics are a
function of routing class– Allows different classes of traffic to take different routes
Packet Routing Strategies
Define Backbone Properties
Backbone Properties
Backbone Routing Protocols
Shortest Measured Delay Routing Parameters
Shortest Measured Delay
The shortest path through the network based on a delay metric where the metric used for each port is the average delay experienced by packets transmitted on the port
The packet delay measurement covers the interval from when a packet enters an output buffer until the packet completes transmission (I.e. transmission queue)
The shortest path calculation is performed periodically and after each node/link failure or restoration causing the routing tables to be updated according to the most recent delay measurements
Vulnerable to oscillations if used with datagram routing– The route that was good for one interval can look bad after the next
interval because too much traffic used it– Can use % deviation and multipath routing to reduce this effect– Sessions using connection-oriented routing are not as vulnerable
because the sessions typically last longer than routing update interval
Common Routing Protocol Parameters
Routing update interval (seconds)– Routing table update interval for adaptive routing algorithms– Updates occur instantaneously in the model– Use traffic sources on the node in order to model broadcasting of
routing table updates
Deviation % for multiple shortest paths– All routes having a metric within X% of the best route’s metric are
regarded as equivalent and traffic is load balanced among the equivalent routes
– When X=0, no ties are declared and only the best route is chosen
Connection-oriented routing for sessions– Session information packets are transmitted over virtual circuits
Packet Routing Classes
Add New Packet Routing Class
Packet Routing Class Parameters
Packet Routing Class Parameters (continued)
Hop Limit– Maximum number of hops that traffic of this class can take between
source and destination
Session Retry Interval (minutes)– Invoked if session setup packet is blocked or lost enroute– Another session setup packet is generated and transmitted after this
delay
Reroute Connections– Connection-oriented sessions routed across links that fail are
automatically rerouted
IGRP Metric– Composite metric used for IGRP routing protocol only– These parameters only apply to traffic classes that are routed with
the IGRP protocol
Session Source Routing Class
Edit Routing Classes
RIP Minimum Hop Routing
RIP Minimum Hop
Routing Information Protocol– Routing algorithm commonly used with TCP/IP– Known as a “distance-vector” algorithm
Treats all links as equivalent and minimizes the number of hops in an end-to-end path
Shortest path calculation performed globally at start of simulation and after each node/link failure or restoration
Hop limits are enforced both in the shortest path calculation and also when routing individual packets
True RIP does not support multipath routing - the % deviation is provided only as a modeling convenience
Link State Shortest-Path-First Routing
Link State Shortest-Path-First
IS-IS (Intermediate System - Intermediate System) and OSPF (Open Shortest-Path-First) are both Link-State-Path-First routing algorithms
IS-IS is used with OSI nets while OSPF is used with TCP/IP integrated IS-IS can be used with either
Routing metric can vary by link and by routing class
Comnet III model includes an arbitrary penalty or weight that varies by routing class, but does not include a traffic dependent component of the metric (use minimum penalty path routing to introduce a traffic dependent component)
Can prohibit a routing class of traffic from using a link by assigning a negative penalty value
Shortest path calculation is performed at start of simulati9n and whenever there is a change in node or link states
Typically much smaller routing update traffic than with RIP
Edit Penalty for Routing Class
Accept Penalty Table
Opposing Port Packet Routing Penalty Table
Minimum Penalty Routing
Minimum Penalty Path
Shortest path through the network based on variable penalties where each port is assigned an arbitrary penalty table
The penalty table assigned to a link port determines the penalty value which can vary by routing class and by the level of congestion on the link as well
Can prohibit a routing class of traffic from using a link for certain (or all) congestion levels by assigning a negative penalty value
The shortest path calculation is performed periodically and after every failure/restoration event using the current metrics assigned to a port
Each port points to a penalty table, which allows different nodes connected to the same link to see different penalties for the link
The model has a list of penalty tables -- typically a different table for different categories of links (e.g. more desirable links have lower penalty values)
Edit Congestion Threshold
Edit Routing Class Penalty
IGRP Routing
IGRP Routing Parameters
IGRP Model Interior Gateway Routing Protocol - A model of Cisco’s routing protocol
Uses a composite routing metric[ K1 * (1010/bandwidth) ] +[ K2 * (1010/bandwidth)/( 256 - (utilization * 255) ) ] +[ K3 * (delay factor) ]
Routing tables are updated periodically (typically every 90 seconds) using current value of metric
Weights K1, K2, K3 and delay factors can vary by routing class. In practice, all weight is often placed on the static components of the composite metric (i.e. K2=0)
Computing the routing metric– bandwidth (bps) is determined by the link in the path that has the minimum bandwidth– utilization is variable– delay factor (units of 10 microseconds) is determined by standard delay parameters
in routing strategy detail by default or by penalty tables if the penalty applied to the port is greater than 1
IGRP uses multipath routing; typically use a very small deviation %
Links
Link Characteristics
Messages are transported– between Processing Nodes and Groups– using Routing Nodes and Switches between LANs
Two links cannot directly connect– Node is required
Takes a message in the form of bytes or packets– reformats it into frames– transfers it to destination node– gives the destination the message in the form of bytes or packets
Proper frame sizes and overheads are built in to the predefined implementations– destination headers on message packets, message checksums, etc.
Link Loading
Frame Parameters
Frame min (bytes)– Minimum size link-level transmission frame– Frames will be padded in order to reach the minimum frame size– Default of zero implies no minimum
Frame max (bytes)– Largest link-level transmission unit– Maximum number of packet bytes = frame maximum - frame overhead– Default of zero implies no maximum
Frame overhead (bytes)– Link-level overhead bytes added to frames prior to comparing the frame to the
minimum frame size– Fractions of a full frame will still incur the full frame overhead
Frame Assembly– Enables assembly of packets from the output queue for a link into a frame
Link Level Transmission Frame
Transmission Frames
PE PE PEProcessingElements
Frames Frames
ApplicationSourceQueue
LocalStoragefor Files
Application Sources & Nodes
Library Command Types
Create Transport Message Command
Message Text Options
Create Setup Session Command
Session Messages
Destination Types
Create Answer Message Command
Accept Command Repertoire
Command Name Detail
Editing “Local” Host Node Commands
Close Local Command Repertoire
Application & Traffic Sources
Process Command Name
Processor Utilization
Accept Process Command Properties
Create Read File Command
File to Access (continued)
Accept Read Command Properties
Accept Command Repertoire
Accept Node Parameter Set
File Size
Create Write File Command
Accept Write File Command Properties
Edit Answer Command
Create Wait For Command
Stop Waiting
Waiting for User-defined State Dependency
Variable Assignment Types
Add New Variable
Accept List of Local Variables
Assignment Expression
Command Name Detail
Macro Command Sequence
Node Command Repertoire
Node Software Capability
Global Commands and Variables
Exercise 3 - Client/Server Network
Replace the Client Session and Server Response Traffic Sources of Exercise 2 with Application Sources and Commands.
Client Workstation Application
In addition to E-Mail, the client workstations still continue to generate a 1000 byte FTP request to the server with an exponentially distributed iteration time (mean of 10 minutes).
The client workstations then wait for a response before they continue processing.
When the server response arrives, the client workstations process the information for 2 seconds and then write a file to a local storage disk. The file size is equal to the server’s answer message size(i.e. A x msg bytes + B).
Try making a global command macro with this sequence of commands.
Execute the global macro in the client application sources.
Server Application
The 100 microsecond server (i.e. processing time per cycle), upon receipt of an FTP request, processes for a random amount of time. The processing time is beta distributed with a minimum of 10000 cycles and a maximum of 40000 cycles. Using the formula (processing time per cycle) * (cycles) = processing time, what does this processing time work out to be on the server?
The server then reads a random sized file from local storage. The size of the file is Poisson distributed with a mean of 10000 bytes.(Hint: Use Poi(10000)).
Finally, the server returns the file to the client with an answer message that is equal in size to the file just read (i.e. A x file bytes + B).
Try making a local command macro with this sequence of commands.
Execute the local macro in the server application source when an FTP request arrives.
Client/Server Node Parameter Sets
Modify the node parameter set for the client workstations so that they have storage disks with 16 millisecond seek times, 200 microsecond transfer rates per sector, and sector sizes of 2048 bytes (i.e. 2 Kbytes).
Modify the server node parameter set so that it has a storage disk with a 10 millisecond seek time, a 100 microsecond transfer rate per sector and a sector size of 16,384 bytes (i.e. 16 Kbytes).
In addition, the processing time per cycle for the server must be set to 100 microseconds.