Transcript CSC 311 - School of Computing Homepage

CSC 311
CHAPTER TEN
CONNECTING NETWORKS
CSC 311
We have looked at several different network topologies
Why do we have different types of networks?
Why do we create different, distinct LAN arrangements?
How do we connect them, so LANs can communicate
with other LANs and WANs?
CSC 311
How we connect networks depends, to some degree, upon
what we are trying to accomplish?
Are we connecting segments of a LAN, similar LANs, dissimilar LANs,
or wide area networks?
CSC 311
Contrary to what the author says, repeaters do not connect LANs, they connect
segments of a LAN. The figure shown below would, logically, behave as
a single Local Area Network.
All devices, on any of these so-called LANs lie in the same collision
domain, and thus, on the same LAN.
CSC 311
As we discussed when examining the Ethernet Topology, Ethernet
Networks today commonly use hubs and their star configuration,
nevertheless, they constitute a single LAN
CSC 311
We would, more likely, wish to use layer two connections, why?
For this purpose, we might use a bridge rather than a repeater.
To use a bridge, all of these LANs must be of the same type.
A bridge does not alter the frames it transmits, all networks would
be using the same MAC protocol. It is not technically correct to refer
to a device connecting two dissimilar LANs as a bridge, it is more of
a router. It does alter frames and must have the MAC protocol of both
types of networks.
CSC 311
How does a bridge know when, or where, to forward a particular
frame, or whether it should be forwarded at all?
Next slide….
Bridges use routing tables, these can be installed on the bridge
by the systems manager or it can be a transparent bridge that will
“learn” its routing table based on the traffic.
following slide…
CSC 311
Routing tables for the network in previous figure.
CSC 311
18. 10 pts. Indicate the contents of the bridge forwarding table given the following
traffic. Indicate after each message, what action the bridge would take and the
current contents of the routing table.
A
B
C
F
G
D
E
1. A sends a packet to C
2 . B sends a packet to E
3. E sends a packet to C
4. C sends a packet to E
5. C sends a packet to D
6. G sends a packet to A
7. D sends a packet to F
CSC 311
You must, however, be careful if you are using transparent bridges,
certain configurations can cause difficulty.
What would the two routing tables look like if B sent a packet to A?
CSC 311
Often there may be multiple paths with different costs associated as
shown in the figure below.
In the absence of some learning protocol, this network would be
flooded with copies of frames
CSC 311
To solve this problem, we use a spanning tree algorithm.
From your data structures: A spanning tree corresponds to a minimal
subset of edges taken from a connected graph that connects the graph’s
vertices.
We must assign a cost associated with each bridge to LAN connection, we
can then produce the following graph.
CSC 311
CSC 311
We won’t go into the details of how to use the spanning tree algorithm.
We would produce the following spanning tree.
CSC 311
This would produce the following topology for our network,
the bridges shown connected by dotted lines are “blocking” bridges,
they will not forward packets.
CSC 311
We have discussed two routing strategies for bridges, namely:
Routing tables loaded into each bridge by the system administrator
FIXED ROUTING
Transparent bridges that “learn” the routes
ADAPTIVE ROUTING, ROUTE LEARNING, ETC.
A third method is SOURCE ROUTING:
The source specifies the route to the destination, routing info
is included in the packet.
CSC 311
Switches and Switched Ethernet
Since Ethernet is the dominant LAN, we will assume, hereafter, that
all of our LANs are Ethernets.
Switches perform the same function as bridges.
While bridges typically connect a couple of LANs, switches may
have a couple of dozen ports which provides many more connections.
Note, that there
is only one switch
in this drawing,
the remaining
devices are
hubs, with no
switching logic
CSC 311
A more common installation, one that is much more efficient, is a
FULLY SWITCHED ETHERNET
Note: there are no collisions in this topology so CSMA/CD is not needed.
CSC 311
Virtual Lans
Use the “intelligence” of the switch to create Virtual LANs that are
really part of a single physical LAN. You might use this approach
to isolate different workgroups working on different projects
without having to physically separate them on separate LANs.
CSC 311
Layer Three Connections
LANs and the layer 1 and 2 connections typically cover
small geographic areas
Wide Area Networks (WAN) span the globe and need more
sophisticated techniques
Much like the highway system, internets or the Internet provide
multiple paths from one point in the Internet to another. We
need a technique to select the best among these many
possibilities.
In the highway example, what do you do if a bridge is out?
Find an alternate route.
Networks are frequently faced with the same problem when one
of the nodes in the network fails or becomes congested.
CSC 311
Generalized Network Topology …. many routes
CSC 311
In order to determine the best route through a network, we need some
way to evaluate those routes. We generally associate a cost with each
link in the route… such as shown below.
Routing tables can be developed using this information to specify
the optimum route between any two points in the network.
CSC 311
Routing Tables may be used. These tables do not usually contain
the entire route to a destination, rather, the next node in the route
to that ultimate destination
This is a partial routing table for nodes A, B, and E
Who defines these routing tables and how?
The process by which a routing table is defined is called a routing algorithm
CSC 311
Routing Strategies
•Centralized
•Distributed
•Static
•Adaptive
CSC 311
Centralized Routing
All interconnection information is generated and maintained at
a centralized location.
That location then broadcasts this information to all network nodes
so that each may define its own routing tables.
One way… routing matrix
Routing Matrix for Network in Figure 10.19
CSC 311
Distributed Routing:
No centralized control.
Each node determines and maintains its routing information
Nodes exchange control information
CSC 311
Static Routing:
Once a node determines its routing table, the node does not change it.
Adaptive Routing:
Nodes respond to changes in the network and update their routing
tables accordingly.
CSC 311
DIJKSTRA’S ALGORITHM
Used to determine path and cost for shortest possible path
between any two nodes in the graph.
The example that we will work will produce the path and the
cost of the shortest possible path between one node and
every other node in the graph.
How does
it work?
CSC 311
Graph for Dijkstra example:
CSC 311
Skip Bellman-Ford
CSC 311
Additional Routing Methods:
Link state routing:
• Each node communicates what it knows to its neighbors
1. bit rate
2. delay
3. queue length
4. reliability
• a node builds a link state packet for each link
• nodes receiving link state packets forward them to neighbors
• as these packets are exchanged , nodes learns about the network
CSC 311
Hierarchical Routing
•Nodes are divided into groups, called domains
•routes between two nodes in the domain are determine by
than networks protocols
•each domain has one or more designated routers
•a large domain may have multiple subdomains.
The “Internet” uses such a hierarchical addressing scheme.
CSC 311
Large networks, such the Internet, can be viewed as a collection of
domains.
In such networks, we define two main categories of routing strategies:
interior and exterior
Interior routing protocols control routing among routers within an
autonomous system or domain
Exterior routing protocols control routers among routers from
different autonomous systems.
To send a
packet from X to
A, we would
use an interior
protocol
-----------------------------To send a packet from Z
to X, Z sends to its router
W, which sends across
subdomains to C, which
sends to A, which sends
to X
CSC 311
This domain concept can be represented using an hierarchical structure
CSC 311
IPv4 Internet Addressing
Internet addresses are 32 bit numbers, usually shown in dotted decimal
notation
143.200.128.3 is an example, each of the decimal numbers can be
represented by 8 bits.
The Internet uses a class address system, each such address can be
interpreted as having two parts: an Internet Protocol (IP) address which
is assigned to the network and a local device address.
The address above is an example of a class B address, the first 16 bits
(143.200), is the IP network address and the other 16 bits (128.3)
represents the address of the particular device on that network.
A class B address provides addresses for over 65,000 possible users,
so it is quite common to further subdivide these users into subnets.
Using subnet addressing, the local management might designate part of
the lower 16 bits to represent a subnet address.
CSC 311
One example of an interior protocol:
Routing Information Protocol RIP
•Use hop count
Hop count from N1 to N2 is 1
Hop count from N1 to N4 is 2
The routers exchange information packets to obtain this information
CSC 311
Another interior routing protocol:
Open Shortest Path First (OSPF)
In this case, the “open” means the algorithm is not proprietary
can also use other factors in determining best route such as delay, bit rates,
actual dollar cost.
CSC 311
Exterior Protocols
Border Gateway Protocol:
current version is BGP-4
used in the Internet to establish paths among routers in different
autonomous systems.
Often concerned with just getting to destination, not necessarily the cheapest
route. There may be other considerations, such as avoiding a particular
country in international traffic.
CSC 311
Within AS3, an interior routing protocol would be used. In leaving AS3, an exterior
routing protocol would be used. The gateway routers (A,B,C,F) , would share
information about their subnetworks.
CSC 311
Congestion and Deadlock:
We have already discussed congestion and flow control. The routing
strategies must deal with the problem of congestion.
As congestion increases the efficiency of the network declines:
How to deal with it?
•Packet elimination
•Flow control
•Buffer allocation
•Choke packets
CSC 311
In the worst case, Deadlock can occur.
Deadlock: congestion is so severe that nothing moves.
Problem described is “store and forward deadlock”
CSC 311
Situation illustrated above is reassembly deadlock:
Buffers are full, packet 0 from both A and B cannot be accepted.
CSC 311
Deadlock prevention:
Reassembly deadlock can be prevented by establishing and reserving
sufficient buffer space to handle the window size for a particular
communication
Store and forward is handled by allocating sufficient buffer space,
but… how much buffer space is enough?
One approach to deadlock is to allow it to occur and then deal with it.
It might be more cost effective to either prevent deadlock or take steps
to reduce the probability of occurrence.
CSC 311
Another deadlock prevention technique uses hop count.
Nodes divide their buffer space into groups based on number of hops,
hop count on a packet is incremented after each hop, a packet must wait
on available space in its hop count group at the previous node, so a packet is
always waiting on a higher numbered group.
CSC 311
CSC 311