Internetworking - Internet Address Depletion and CIDR Introduction

Document Sample
Internetworking - Internet Address Depletion and CIDR Introduction Powered By Docstoc
                                                ►A  subnet is a subset of class A, B, or C
   Internet Address Depletion
                                                ► IP addresses are formed of a network and
            and CIDR                              host portions – network mask used to
                                                  separate the information

              Introduction                                       Introduction
► Each class of address has its own “natural    ► Mask of is applied to network
 mask” – mask created by the definition of
 the network
   class A natural mask                    divides the IP address into a network
   class B natural mask                  portion of 10, subnet portion of 0, host portion
   class C natural mask                of 0.1
► By using masks, networks can be divided
 into subnetworks
   extends the network portion of the address
   into host portion
   increases the number of subnetworks and
   reduces the number of hosts

 Variable Length Subnet Mask                                 Variable Length Subnet Mask
► VLSM  allows a network to be be
 configured with different masks
   adds more flexibility in dividing the network
   into multiple subnets
   without VLSM a mask may have too few
   subnets or hosts
► Suppose  we want to split
 (class C) into three subnets with 100 hosts
 in one subnet and 50 hosts in each
 remaining subnet

                      CIDR                                                      CIDR
          Inter-Domain Routing was
► Classless                                                 ► Classless   addresses
 designed as a remedy for                                      main observation: many organizations need
   class B exhaustion                                          more than a class C network but does not
                                                               have enough hosts to efficiently utilize a class
   routing table explosion                                     B
    • as more networks get connected -- more memory            idea: give such organizations multiple class C
      is needed for storing routing tables                     addresses
    • most high performance routers “cache” portions of        in the CIDR strategy, the class C addresses
      routing tables at the interface board themselves --      are contiguous and share the same “most
      to speedup forwarding                                    significant bits” -- the same prefixes
    • some extreme designs had fast memories that              if the routing protocols can route based on
      were in stand-alone mode at the interface boards         these prefixes, they need only one block of
                                                               network numbers

                        CIDR                                                CIDR
       by allocating addresses intelligently -- we can   ► CIDR  notation enables lumping of specific
       group numbers by region                             routes into aggregates
► InCIDR, an IP network is represented by                ► Aggregate denotes any summary route
  a prefix
                                                         ► Supernet denotes a summary route with
       IP address + some indication of the left-most
                                                           shorter prefix length than the natural mask
       contiguous significant bits within this address
►A  network is called “supernet” when prefix
  boundary contains fewer bits than the
  networks natural mask

                        CIDR                                                CIDR
                                                         ► Networks    that are subset of an aggregate
                                                           or a CIDR block are called “more specific”
                                                         ► Routing domains that are CIDR-capable
                                                           are called “classless” – traditional routing
                                                           “classfull” routing

     Route Aggregation in CIDR                             Route Aggregation in CIDR
                                                      ► Aggregation    may not work always
                                                           customers having IP addresses that do not
                                                           belong to their provider’s range
                                                           some customers (ISPs) need to connect to
                                                           multiple providers at the same time
                                                      ►A  router with and
                                              will match when
                                                       trying to deliver traffic to

                                                       Single Homing: Address Outside
          Longest Prefix Match
                                                          Provider’s Address Space
►   Destinations connected to multiple domains        ► Customer  connected to single provider
    must be explicitly announced – in most specific   ► IP address space different from provider’s
                                                      ► Customer changed providers and kept
                                                        addresses of the previous provider
                                                      ► Renumbering should be done – if not
                                                           provider cannot aggregate as efficiently – hole
                                                           is punched in the address space
                                                           new provider cannot aggregate the address

    Multihoming Scenerio: Addresses                         Multihoming Scenerio: Addresses
         taken from one provider                                 taken from one provider
► Customers   are connected to multiple
  providers – small enough to take
  addresses only from one
► Aggregate advertisement can lead to black
► Aggregating someone else’s routes (proxy
  aggregation) can be tricky
      unless aggregating party is a superset
      or parties are in total agreement

    Multihoming Scenerio: Addresses                         Multihoming Scenerio: Addresses
         taken from one provider                                 taken from one provider
►   ISP2 sends an aggregate summarizes Jamesnet
    and Lindanet into one update
►   Stubnet which is a customer for ISP1 has an
    address space falling in
►   Traffic for Stubnet will perform
    longest match and endup in ISP2
►   Solution:
      ISP2 should specifically list each of the IP ranges
      that it has in common with ISP1 on top of its own
      address space

Multihoming Scenerio: Addresses       Multihoming Scenerio: Addresses
     taken from one provider           taken from Different Providers
                                  ►   Large domains can take addresses from
                                      different providers
                                  ►   Each provider aggregates its own address
                                      space without listing specific ranges from other
                                        drawback – backup routes to multihomed
                                        organizations not maintained – redundancy is one of
                                        the reasons for multi-homing!
                                        traffic using the addresses taken from provider will be
                                        unable to reach the destination if the provider is down
                                        – even if the destination is reachable via “other”

Multihoming Scenerio: Addresses
 taken from Different Providers