Multicast - Introduction

Boa noite Pessoal!!!

Vamos agora a uma breve introdução à Multicast. Para os que não conhecem, sugiro que deem atenção, pois é um assunto muito importante para nós profissionais de rede, e é o tipo de assunto que quando aprendemos, não sabemos porque não é mais aproveitado!

Só tem um detalhe. Como a prova do BSCI do CCNP é em inglês somente, claro que o documento será em inglês ;-)

- The range of address is 224.0.0.0 - 239.255.255.255.

• 224.0.0.0/24 is reserved by IANA. They are known as local network control block. The TTL of this range is usually set to 1.
  
• 224.0.1.0 - 238.255.255.255 are called Globally scoped addresses. This range was to be used to enable mbone (multicast backbone), so that multicast could flow over the internet. But this project is almost dead.
  
• 239.0.0.0/8 are known as Limited/Administratively Scopoed IP Multicast.
  

- Hosts use IGMP (Internet Group Management Protocol) to talk to their local routers to register and unregister to multicast groups.


- To verify the groups and users of the groups: router# show ip igmp group

- on IGMPv1, the router sends membership queries to the address 224.0.0.1 (all hosts) every 60 to 120 seconds. Report suppression is used to minimize the multicast traffic, by letting only one host in each subnet responsible to respond the query packets. There is no mechanism for a host to leave a multicast group.

- on IGMPv2, some advantages were implemented:

• Group-specific queries are now sent to specific groups, instead of requesting a report to all groups.
  
• Leave Group messages are sent by the host to the router specifying that they are leaving the group. When the router receives this message, it also asks the other routers on the segment if they still wanna receive messages from an specific group.
  
• Querier election mechanism is used to elect one host in a segment as the designated querier. The host with the highest ip will be elected.
  
• Query-interval response time is included in the query and indicates to the members how much time they have to respond to a query.
  

- on IGMPv3, the major improvement is that the client can specify filtering to be able to listen only to specific sources. The report messages are sent to 224.0.0.22 (IGMP) instead of 224.0.0.2 (All routers).


- To verify which version of IGMP is being used: router# show ip igmp interface interface.


- Since IGMP is a layer 3 protocol, layer 2 switches are not aware of it, and threat multicast just as broadcast. To prevent it, there are two methods:

• CGMP (Cisco Group Management Protocol): The router informs the switch which hosts are joined in a group, and the switch is going to forward the multicast messages only to these hosts.
  
• IGMP Snooping is the act of listen the IGMP traffic and update the MAC address of the multicast address to the ports where the hosts of a group are in. The only outcome is that since the switch will have to listen to the multicast addresses, it may overload the switch.
  

- PIM (Protocol Indepent Multicasting) Routing Protocol is used by the routers to forward the multicast messages. It uses the unicast routing table in its multicast routing calculations. Multicast routing is not preocupied on finding the route to the destination (multicast address), but to the source (client address) using the unicast routing table.
There are two types of distribution trees that PIM uses to forward multicast messages, since it's not like the unicast routing:

• Source tree: an entry is created for each source(S) sending to a multicast group (G), and then it forms a spanning-tree to forward traffic to each multicast address. It will happen for each router in the path.
  
• Shared tree: is a single tree shared by the routers that have a single common root, called Rendezvous Point (RP). The routers will initially send the messages to the RP, and the RP will forward to the destinations.
  

RPF (Reverse Path Forwarding): is an additional check to prevent loop. Basically the router checks if the best route to the subnet that the source host is, is on the same interface that the multicast packet came from.


- PIM uses the following notation:

• (S,G): source(S) sending to group G. Tipically represents a source tree, but may also appear on a shared tree.
  
• (*,G): seen on a shared tree, because in this mode, the router does not case which is the source, it will send to the packet to the RP anyway.
  

- PIM has three modes of operation:

• Dense: It uses a source tree mode, and when a client sends a packet to a multicast address, the router will flood its neighbors with the packet, and the routers that don't want it, will respond back to the sender informing they don't want that packet. To turn it on, you gotta configure it for each interface: router(config-if)# ip pim dense-mode
  
• Sparse: It uses the shared tree mode, that elects a RP. To enable it: router(config-if)# ip pim sparse-mode
  
• Sparse-Dense: If you are unsure what version of PIM the other routers are configured, you can run both with this mode. To enable it: router(config-if)# ip pim sparse-dense mode.
  

- Follows the basic configuration of PIM:

• To configure the address of the RP: router(config)# ip pim rp-address rp-address
  
• To auto-announce you as the RP: router(config)# ip pim send-rp-announce
  
• To auto-discover the RP: router(config)# ip pim send-rp-discovery interface scope. The scope is the number of hops you want to try to discover the RP.