Routing Protocol Fundamentals

  • What parts of enterprise design do you know?
    • Building Access: layer 2, user access to network
    • Building Distribution: this aggregates building access layer, multilayer switches
    • Campus Backbone: high speed transfer of data, high-end multilayer switches
    • Edge Distribution: handle traffic in and out of the Campus network, routers or multilayer switches
    • Internet Gateways:
    • WAN aggregation: remote offices, (ie.: MPLS)enterprise-network

Routing Protocol Selection

  • What is an IGP / EGP? What is the difference? Specify them!
    • Internal Gateway Protocol: exchanges routes between routers inside a single AS. EIGRP, OSPF, IS-IS, RIP, iBGP
    • External Gateway Protocol: exchanges routes between autonomous systems. BGP
  • What are the key factors?
    • Scalability
    • Vendor Interoperability: IEEE vs Cisco vs other
    • IT staff knowledge
    • Speed of Convergence: The amount of time for the failover to occur is the convergence time.
    • Capability to Perform Summarization
  • What is the main logic of Distance Vector routing protocol? Specify the DV routing protocols!
    • Neighbor gossip. Advertises the full routing table. EIGRP, RIP
  • What is a routing loop?
    • If R1 loses one of its route R2 advertises back to R1 R1’s own failed route. Some ping-pong while the metric is incrementing until it is infinite and dropped.
  • How do you prevent routing loop?
    • Split HorizonThe Split Horizon feature prevents a route learned on one interface
      from being advertised back out of that same interface.
    • Poison Reverse: Useful when Split Horizon is not active. Sets the received route metric to infinite if it is going out on the same interface where it arrived.
  • How does RIP work?
    • Uses hop count as metric. Periodic updates. The max hops is 15 ( 16 =< infinite ). RIPv1: broadcasts and only fixed -length subnet masks. RIPv2: +multicast + variable-length subnet masks. RIPng: + IPv6
  • How does EIGRP work?
    • First sends out full routing table then only triggered updates. Converges quickly. Bandwidth + delay + K values as metric. DUAL algorithm.
  • What is the main logic of Link State routing protocol? Specify the LS routing protocols!
    • Link State routing protocols build the whole topological map of the network. Everybody knows every path. Uses Dijkstra Shortest Path First algorithm. Routing updates sent in response to the network changes. OSPF (BR,BDR), IS-IS
  • How does Path vector work?
    • Path Vector protocol knows the exact path packet take to reach a network (through multiple AS).

Network Technology Fundamentals

  • What type of network traffic do you know (4, detailed)?
    • Unicast: from single source to single destination,
    • Broadcast: single source all destination in a subnet, 255.255.255.255, not used in IPv6
    • Multicast: devices wanting to receive a given traffic can join the multicast group, switches and routers can determine which port should they forward the multicast traffic,
    • Anycast: one-to-nearest. With IPv6 one IP address can be assigned to multiple devices and anycast goes for the one which is closest to the router. How? BGP supports having the same network to multiple path.
      This is useful for services like DNS.
  • What network architecture type do you know (3, detailed)?
    • Point-to-Point Network: a single link which connects 2 routers, ie: serial links
    • Broadcast Network: a broadcast sent from a router reaches all other routers  on a segment, ie: ethernet network
    • NBMA: non broadcast multi-access network: no broadcast or multicast messages available, only unicast.
      In addition Split horizon and Designated router selection has problems. (no advertisement on the same interface back, though split horizon can be turned off)
      NBMA

TCP/IP Fundamentals

  • What is route summarization?
    • More routes are summarized into one bigger subnet (still the smallest intersection)
  • How do IPv4 /IPv6, ICMP, TCP and UDP header look like?
  • What is asymmetric routing and when could it happen?
    • For the same source and destination the traffic chooses another path to reply. FHRP could cause this.
  • What is MTU?
    • Maximum Transmission Unit: this is the largest packet size supported by the interface. 1500 bytes is the average. On slower links large MTU can cause latency.
  • What is TCP MSS?
    • TCP Maximum Segment Size: the amount of data can be contained in a single TCP segment. The MMS refers only to the amount of data.
  • What is the relation between MTU and TCP MSS?
    • MTU-image-1.png
    • In the above example the MSS is 1460 bytes.
      The MTU is 1460 + 40 = 1500 bytes. There is no problem here.
      In case we use GRE:
      MTU-image-2
      The MTU will be 1460+40+24 = 1524 bytes.
      This means that fragmentation is needed to transmit the packet. Unfortunatelly fragmentation also mean overhead, as the protocols will use 2* 66 bytes for the 2 packets.As a solution TCP MSS can be adjusted so fragmentation will not be necessary.
      With the same example (GRE): MSS = 1436
      So a packet will consist of  1436+20+20+4+20=1500  Fragmentation is not needed !
  • Which is the most famous ICMP utility?
    • ping, traceroute
  • What is the 2 most important ICMP message type?
    • Destination Unreachable: if the packet enters a router which doesnt know how to reach the target IP then it can reply with ICMP destination unreachable
    • Redirect: The host might have information about the next-hop IP but it is incorrect. The router can redirect the packet to the right next-hop address.
      icmp-redirect
  • What is a three-way handshake?
    • This is the process of setting up a TCP session: SYN -> ACK, SYN -> ACK
      3wayhandshake

TCP Sliding Window

  • How does TCP windowing work?
    • TCP communication sends a packet then it receive an acknowledgement about the arrival. But it can send multiple packets between 2 acknowledgements. The window size is growing exponentially after every ackknowledgement (first).
  • What is TCP slow start and when does it happen?
    • When a single TCP flow drops a packet the window size will be reduced to one segment (TCP slow start). After that the window size will exponentially grow until reaches one-half of its congestion window size. At that point the window size will grow linearly.
  • What happens when the interface’s output queue fills? How do we call this effect?
    • In this case all TCP flow will simultaneously start to drop packets (tail drop) which causes TCP slow start to all of them. This is global synchronization or TCP synchronization.
  • What is the biggest problem with the previous effect?
    • TCP slow start is not a problem but when all TCP flow do it (global sync) the bandwidth is used very inefficient. (gaps).
      global-sync.png
  • How can we prevent this? How does these techniques different from Tail Drop?
    • Tail Drop: If the queue is full the new packets are dropped.
    • Random Early Detection (RED): If the queue is empty no packets will be dropped. The more packets are in the queue the more packets will be dropped. If the queue is full every new packets will be dropped.
    • Weighted RED (WRED): Like RED but it differentiates more and less important packets to drop.
  • What is Out-of-Order delivery?
    • For example there are 2 outgoing interface on a router and some packet using one of them while other packets use the other one so the packets might arrive to the destination in a different order. To prevent this TCP uses sequence numbers for the correct order or the destination device asks for retransmission for the out of order packets.

UDP

  • Why is different UDP from TCP?
    • UDP is connectionless, unreliable protocol without sequence numbers, acknowledgements or window sizes.
  • What kind of traffic is optimal for the use of UDP? Why?
    • In case of real time traffic the best choice is UDP because it is much smaller then TCP (less header) and faster (less communication).
  • What happens if UDP packets arrive out of order?
    • UDP has no sequence number field so it has no idea in which order a segment arrives. It will just pass any segment to the upper layers as they arrive.
  • Which protocol is encapsulated inside UDP (for example voice traffic)?
    • Realtime Transport Protocol RTP: layer 4, the latency ideally smaller than 150ms
  • How can we reduce the latency of this protocol?
    • With Low Latency Queuing (LLQ) one or more traffic types can be buffered in a priority queue which is served first. Works like a fast lane in the highway. There is a bandwidth limit for LLQ so it doesn’t starve out the non-priority traffic.

Network Migration Strategies

  • What are the 2 common techniques for routing protocol migration?
    • Using Administrative Distance: configure both routing protocols, but the existing one will have higher AD. After the new protocol seems to be OK, lower the original routing protocols AD.
    • Using Route Redistribution
  • How can we migrate to IPv6?
    • check if IOS is capable of IPv6
    • Most devices are capable of dual-stack configuration which means both IPv4 and IPv6 can be run at the same time.
    • NAT64: In a case where IPv4 have to communicate with IPv6 we can use NAT which translate IPv6 to IPv4 address and back so they can communicate.
    • Send IPv6 traffic over an IPv4 tunnel
Advertisements