Internet Network Layer Flashcards
What are the 5 responsibilities of internet network layer services?
- Move packets from an origin to a destination node
- Avoid congestion
- Internetworking
- Lowest layer dealing with end to end issues such as routing
- Must know topology to chose eligible paths
What is the difference between congestion and flow control?
- Congestion control: looking over the whole network making sure it can carry all the traffic offered
- Flow control: O-D point to point control, fast-senders shouldn’t overwhelm slow receivers.
What is the difference between congestion and flow control?
- Congestion control: looking over the whole network making sure it can carry all the traffic offered
- Flow control: O-D point to point control, fast-senders shouldn’t overwhelm slow receivers.
What 2 things are specific to congestion control?
- Under heavy traffic the performance will eventually collapse
- All resources (node capacity, service discipline) are involved.
What is a connection oriented network layer service? and describe it?
Virtual Circuit (VC):
- Routing decision only set up once (at the set-up time)
- Nodes know where to forward the packets
What is a connectionless network layer service?
Using Datagrams:
- No route is given in advance (each datagram knows full destination address)
- Datagrams can follow different routes
Pros and Cons of using datagrams oppose to connection oriented service?
Pro: More robust and can cope with congestion
Con: More work to setup i.e overhead
How does data flow in virtual circuits?
Data flow begins and arrives at the destination node in sequence all along same path.
How does data flow in datagram networks?
Datagram 1 , datagram 2 and 3 may not all take the same path and they may not all arrive at dest node in correct order.
What happens to a datagram during routing failure compared to to a VC?
Datagram: The datagram passing through that router is lost but not necessarily the others.
VC: All VCs that passed through the failed router are terminated
What state information do routers hold in datagram vs. VC subnets?
Datagram: Routers don’t hold state info about connections
VC: Each VC requires router table space per connection to know where to forward to according to path
What are the 3 different types of protocol prevalent in the internet network layer?
- IP protocol
- ICMP protocol
- Routing protocols
What 3 things are involved generally in IP protocols?
- Addressing conventions
- Datagram format
- Packet handling conventions
What 2 things are involved generally in ICMP protocols?
- Error reporting
- router signaling
What 3 things are involved generally in routing protocols?
- path selection
- forwarding tables
-RIP, OSPF
What does a routing algorithm do?
Determines end-to-end path through network
What does a forwarding table in a router do?
Forwardring table determines local forwarding at this route
Where is the routing algorithm kept and where is the forwarding table kept?
Routing: In the control plane
Forwarding: In the data plane
What is the control plane?
How datagrams are routed on an end-to-end path from sender host to destination host
Where is the control plane usually implemented?
Remote controllers
What is the data plane?
How datagrams arriving to an input link is forwarded to the router’s output link.
What are the 4 components of a router’s architecture?
- Input ports
- Output ports
- high-speed switching fabric (data plane; hardware)
- routing processor (control plane; software)
What is WFQ? Describe it also
Weighted-Fair-Queuing:
- Generalised round robin
- Each traffic class gets weighted amount of service in each cycle
Can routers fragment datagrams?
Yes a large datagram at input can be fragmented into several smaller ones to then be reassembled at the destination
Can routers fragment datagrams?
Yes a large datagram at input can be fragmented into several smaller ones to then be reassembled at the destination
What does CIDR stand for?
Classless inter-domain routing
What is classless inter domain routing?
- There is a subnet portion of arbitrary length
- The address format is a.b.c.d/x where x is the number of bits in the subnet portion of the address
- The subnet portion length can change based on num of hosts attached
Why is CIDR used?
Because it allows many users from the same network to hvae an IP address that is recognisably similar to each other - aka. they all belong to same subnetwork
What does DHCP stand for?
Dynamic host configuration protocol
What three things does a host need to connect to the internet? And how often do these things change
- An IP address
- A subnet mask
- Address of nearby router
- Each time a user moves or relocates the elements must be reconfigured
Explain how an incoming client would be given an IP address in a network?
- Incoming client sends request to join network
- DHCP server sends back:
- an unused IP address
- Address of 1st hop router for client
- Name and IP address of DNS server (domain name system)
- The network mask
How did networks transition from IPV4 to IPV6?
The entire IPV6 datagram including header fields, source/dest is held as the payload in the IPV4 datagram.
What is tunneling?
When an IPV4 tunnel connects two IPV6 routers
How does tunneling between IPV6 routers along an IPV4 tunnel work?
- At IPV6 router before tunner - the datagram is encapuslated in IPV4 datagram and the Source and Destination are changed from current source (aka current IPV6 router) to the next IPV6 router after the tunnel
- When it reaches that next IPV6 router after tunnel it uncaspulates and then can go forward to destination IPV6 routers
What are 2 types of routing algorithms?
Non Adaptive
Adaptive
What are non-adaptive algorithms? (1)
- No consideration to traffic/topology changes (e.g static routing)
What are adaptive algorithms? (4)
-React to changes: Topology and traffic patterns
What are 3 types of adaptive algorithms?
- Centralised (Global Information)
- Isolated (Local Information)
- Distributed (Global and Local iformation)
What are 2 types of least cost algorithms for networks?
Distance vector
Link State algorithm
What is the distance vector an example of?
Decentralised routing algorithm
Describe the distance vector algorithm?
- Calculates the least-cost-path in an ITERATIVE and DISTRIBUTED manner.
- Each node begins with only the knowledge of the costs of its attached links
- Then by exchanging information with neighbour nodes, nodes gradually calculateleast-cost path
What is an example of a distance-vector algorithm?
Bellman-Ford’s Algorithm
Do an example of the distance-vector algorithm
Remember:
- Estimate of least cost from x to y is written Dx(y)
- When you don’t have the cost of one node to another you mark its cost as INFINITY
Describe in 3 steps the distance vector algorithm?
- Wait for a change in local link cost or a message from neighbours.
E.G c(x,v) (cost of x to v) - Update distance estimates for desired distance
- If distance estimates have changed - notify neighbour
What is the link state algorithm an example of?
A global routing algorithm
What is meant by a global routing algorithm?
- Calculate least-cost path using complete, global knowledge about the network
- Need mechanism to obtain this information before calculations (flooding)
- Calculations can be performed on site or replicated at multiple sites.
- Referred to as link-state algorithms
Example of global routing algorithm?
Dijkstra’s algorithm
What is message complexity of LS compared to DV routing algorithms?
LS: n nodes, E links, O(nE) messages sent
DV: exchange between neighbours only
What is speed of convergence of LS compared to DV routing algorithm?
LS: O(n^2) algorithm requires O(NE) messages –> these may have oscillations
DV: may have routing loops (aka count to infinity problem)
Comparison of robustness to router malfunction of LS and DV?
LS:
- Node can advertise incorrect link cost
- each node computes its own table
- Errors propagate less
DV:
- DV node can advertise incorrect path cost
- Each node table is used by others so ERRORS propagate
How are autonomous systems connected to each other?
Autonomous systems can be connected to each other via EGP
- Relies on AS numbers to construct AS paths
- exterior gateway protocol routing algorithms
- These are carried out by the edge connectors
How do autonomous systems route internally?
Using IGP (interior gateway protocol routing algorithms)
- all routers in an autonomous subsystem engage with this
- Relies on IP addresses to construct paths
What is an example of EGP?
BGP (border gate protocol)
What is an example of IGP?
RIP (routing information protocol), OSPF (open shortest path first)
Where do routers using RIP learn information from? And how often is it updated
- In RIP each router learns the distance to each destination from its neighbours
- Router sends update message to its neighbours every 30 seconds
What is the metric to compute shortest path for RIP?
Number of hops
2 Facts about RIP?
- Runs on top of UDP
- Uses mechanisms to reduce routing loops
2 facts about OSPF?
- Runs over IP (link-state)
- OSPF enables each router to learn about the complete network topology.
What does each router do in OSPF?
- Monitors the cost of the link to each of its neighbours
- Floods the link-state information to other routers on the network
What is the inter-domain routing behind BGP based on?
Classless address prefixes as well as policy-based routing
3 facts about BGP?
- BGP establishes TCP connections
- BGPv4 is a de-facto inter-domain routing protocol
- Provides mechanisms for address aggregation
What do the routers using BGP keep?
A global view of the internet in their RIBs (routing information bases)
What does a BGP speaker do?
Exchange network reachability information with other BGP speakers (peers) adjacent to it –> (CONTAINING sequence of ASs that packets must traverse to reach a destination network)
What are 2 famous internet management protocols?
- ICMP (internet control message protocol)
- SNMP (simple network management tool)
What does network management include? (2)
- Deployment (integration and coordination of HW and SW and human elements)
- Monitor, test, poll, configure, analyse, evaluate and control the network and element resources to meet operational requirements
Can network agents be hardware?
Yes, network agents can be hardware or software
Describe the network management framework?
- Managing server, (MS) with personnel
- Managed device (MD) equipment (HW+SW)
- MO: HW configuration parameters (e.g AS routing OSPF)
- MIB (management information base)
- Network management agent (inside the MIB)
- Network management protocol
Where is network management protocol needed?
Between the managed device and the managing serer
What is SNMP briefly?
- Application layer protocol to convey network-management control information messages between managing server and agent
- It lies on top of UDP
What is the most common use of SNMP?
Request-response mode: managing server send a request to an SNMP agent who performs some actions and sends a reply to request
What is an example of an SNMP managing server request?
modify MIB (management information base) values.
What is ICMP briefly?
- Used by hosts and routers to communicate network-layer information to each other e.g error REPORTING
- Lies on top of IP (it is not part of IP)
- ICMP messages are carried in IP datagrams as payloads
Example of the use of ICMP
When you get a message back form a HTTP session saying you can’t reach a destination network or something - that message is an ICMP message
How does a host read an ICMP message delivered in an IP datagram?
It demulitplexes the datagram content to ICMP
