Module 3: Introduction to Physical, Data and Network Layer Flashcards
Switch
mechanism to connect links to form a larger network
multi-input, multi-output which transfers packets from an input to one or more outputs
physical layer: electrical signals
link layer: frames
network layer: packets (routers)
physical layer: repeaters
distance limitation
- electrical signal becomes weaker as it travels
- imposes a limit on the length of a LAN
repeaters join LANs together
- analog electronic device
- continuously monitors electrical signals on each LAN
- transmits an amplified copy
physical layer: hubs
multi-port repeater
joins multiple devices electrically
- do not necessarily amplify the signal
- broadcast traffic that arrives on every other port
very similar to repeaters
- also operates at the physical layer
limitations of repeaters and hubs
one large collision domain
cannot support multiple LAN technologies
ethernet: utilization > 30% is considered heavy due to collisions
what is a solution to the distance problem with switches and routers?
link layer bridges
- connects two or more LANs at the link layer
link layer switches
typically connects individual computers
- a switch is essentially the same thing as a bridge
- … though typically used to connect hosts, not LANs
like bridges, support concurrent communication
- e.g. host A can talk to C while B talks to D
what is the main idea of bridges/switches?
main idea is traffic isolation
bridge/switch breaks network into LAN segments
bridge/switch filters packets
- frame only forwarded to the necessary segments
- segments become separate collision domains
what are the advantages of bridges/switches over hubs/repeaters?
- only forwards frames as needed
- filters frames to avoid unnecessary load on segments
- sends frames only to segments that need to see them
- separate collision domains allow longer distances - joins segments using different technologies
- e.g. ethernet - made possible with buffer
- not complete heterogeneity
what are the disadvantages of bridges/switches compared to hubs/repeaters?
delay in forwarding frames
- store and forward: bridge/switch must receive and parse the frame and perform a lookup to decide where to forward
- solution: cut-through switching
–> look up destination address in header, determine output port, forward frame
–> no error detection
higher cost!
- more complicated (intelligence, memory buffers, high performance) devices that cost more money
why are switches so expensive?
high performance chips
memory requirements
intelligence capability makes it complex to build
bridge/switch forwarding table
protocol
self learning: building the table
self-learning: without intervention from network admins
when a frame arrives:
- inspect the source MAC address
- associate the address with the incoming interface
- store the mapping in the switch table
- use a time-to-live field to eventually forget the mapping (in case host moves to another network)
when frame arrives with unfamiliar destination
- forward frame out all of the interfaces
- except for the one where the frame arrived
- hopefully, this case won’t happen very often
flooding can lead to loops
flooding can lead to forwarding loops
- if the network contains a cycle of bridges
- either accidentally or by design for higher reliability
what is the solution for avoiding cycles?
spanning trees
ensure the topology has no loops
- avoid using some of the links when flooding
- … to avoid forming a loop
spanning tree
- sub-graph that covers all vertices but contains no cycles
- links not in the spanning tree do not forward frames
spanning tree algorithm
steps in spanning tree algorithm
robust spanning tree algorithm
we must react to failures
can we deploy bridges/switches to the scale of the internet?
this is a layer 2 technology
-size of bridge forwarding tables grows with the number of hosts
-broadcast traffic flows freely over the entire extended LAN
-little control over forwarding –> send along spanning tree –> too much flooding
-spanning tree topology: hotspots, single point of failure, poor load balancing
-limited in heterogeneity: ethernets to ethernets, ethernets not to ATM
Physical LAN / virtual LAN
physical
all ppl in adjacent officers were put on the same LAN
independent of whether they belonged together or not
virtual LAN
single extended LAN partitioned into several separate LANs
independent logical networks (VLANs) within the same physical network
configured in software rather than hardware
advantages of VLANs
group users based on organized structure rather than physical layout
rewire the building in software
isolating traffic on separate LANs by organization and improves security and performance
example VLAN
configuring VLANs
using the graphic, explain how VLANs help avoid flooding the network.
Network Layer
Internetwork
Internet Protocol
IP Header
fragmentation issues
different networks may have different frame limits (MTUs)
iPv4: fragment on demand and reassemble at destination
iPv6: network returns error message so host can learn limit
example fragmentation issues
reassemble at destination (H8)
fragmentation and resassembly
this can lead to problems along the way
we could lose one of the fragments
how does the IP header support fragmentation?
fragmentation fields
explain the fragmenting of a package
what are some fragment considerations?
other considerations for fragmenting
causes inefficient use of resources within the network
higher level protocols must re-xmit entire datagram (on lossy network links, hard for packet to survive)
efficient reassembly is hard
how does IPv6 do fragments better?
how does IPv4 approach fragmentation?
fragment on demand and then reassemble
why is the fragmentation approach of IPv6 better?
And you wouldn’t do this if you just have one packet to send. But suppose you have set up a large flow and you want to send a lot of data, then using this PMTU approach, you can probe the network on the hop by hop basis, figure out what the MTU is, and use that right from the beginning. And that gets around all the complexity of reassembly, what if you lose a fragment, and so on. Consequently, this is a much better approach compared to IPv4.
virtual machines vs. containers
