Data communication and network-IGMP

Document directory

• Group Management
• IGMP operation
• Encapsulation

Internet Group Management Protocol (IGMP) is an auxiliary protocol for IP multicast. So what is multicast?

"Multicast" is also called "multi-point transmission" (Multicast). A package sent by a host can be received by multiple other qualified hosts at the same time, this host and those qualified hosts form a group, and their communication within the group is broadcast-type. The principle of multicast is to set the NIC of some hosts on a network to multicast transmission mode, specify that it does not filter data frames with a multicast transfer address as the destination physical address. In this way, the drivers of these hosts can receive data frames with the multicast transfer address as the destination physical address at the same time, but the drivers of other hosts cannot, these Hosts form a "Multicast" group logically. This technology can effectively reduce the burden on hosts other than multicast groups on the network, because the data sent to the multicast group will not be transmitted to their drivers for processing, avoiding unnecessary waste of resources. At the beginning, the purpose of designing this technology was to make up for the lack of broadcasting communication.

In fact, multicast or broadcast (because the switch cannot find the physical destination address of multicast frames, the physical destination address of multicast frames will be discussed later in this article ), only machines that do not join the multicast group have filtered out the multicast at the hardware layer, reducing the waste of computing resources on the machine, but still wasting network resources, therefore, Cisco has developed the cGMP Cisco Group Membership Protocol Cisco group membership agreement. cGMP works in the Client/Server mode. The vro is equivalent to a server, and the vswitch is equivalent to a client. After the vro receives the IGMP message, notifies the vswitch multicast receiver to join or exit the multicast group using the cGMP command, and instructs the vswitch to perform accurate multicast-targeted forwarding.

Let's go back and talk about IGMP ~

For multicast on the English site, group management requires vrouters capable of routing multicast groups. The route tables of these vrouters must be updated by some multicast routing protocols, these multicast routing protocols will be described later. IGMP is not a multicast routing protocol, but a group membership protocol. In any network, one or more multicast routers distribute multicast groups to hosts or other routers. The IGMP protocol is used as a multicast router (Multicast router ), provides information about the status of host (router) members connected to the network. There are two versions of IGMP in total. This article discusses the igmpv2igmp operation. For each group, a vro is responsible for distributing multicast groups to the specified group, if there are three multicast routers connected to the network, their group identifiers are mutually exclusive. Only vror R distributes multicast packets whose IP address is 239.2.59.1. Vror R is a distribution router. R1 and R2 depend on R to maintain the group list. They are the recipients of R in this network, but for other networks, r1 and R2 may be distribution routers, but they are not in the current network. A host or vro can be added to a group. Each host maintains a table of members in the group. When a process is to be added to a new group, it sends a request to the host, the host adds the name of the process and the name of the requested group to its table. If this is the first member relationship request in the group, the host sends a member relationship report packet to the multicast router (the address is the host address, all members in this CIDR block can receive the report, and other Members will no longer send the report packet after receiving the report.) If this is not the first member relationship, no member relationship message needs to be sent, because the host is already a member of the group, it has received multicast groups for this group. When a host discovers that no process has been added to a special group, it sends an exit report. Similarly, when the vro finds that there is no interface connected to the network to add a request to a special group, it finds a departure report about the group (Address: 224.0.0.2, all multicast routers ). However, when a multicast router receives an exit report, because the report only comes from one host or route, other hosts or routers may still have requests to join the group, therefore, it is impossible for him to immediately clear the group from its multicast address table. To ensure this, the vro sends a special query packet and inserts a Group Identifier or multicast address related to the group, the vro allows a specified time to respond to the host or vro. If no join request is received within the specified time, the router assumes that there are no loyal members in the network and deletes the group from its list. A member relationship host or router can be added to a group by sending a member relationship report message, or you can leave a group by sending an exit report, however, these two types of reports are insufficient. In this case, a host adds a request to a group, but the host is closed or deleted from the system. Multicast routers will never receive outgoing packets. Therefore, the vro periodically sends common Query Packets (Address: 224.0.0.1, all hosts). When the host or vro receives a common query packet, if it has a request to join a group, you can use a membership report to respond. The encapsulated IGMP packet is encapsulated in the IP datagram, while the IP datagram itself is encapsulated in the frame. For the IGMP protocol encapsulated at the network layer, the protocol field of the IP Group is 2. Each IP group carrying this value in the protocol field needs to transmit data to the IGMP protocol, when the packet is encapsulated in an IP datagram, the TTL value is set to 1. Because the IGMP domain is in the LAN, IGMP packets cannot be transmitted outside the LAN. This TTL ensures that messages cannot be sent from the LAN. The data link layer is encapsulated at the network layer. IGMP packets are encapsulated in IP groups and processed as IP groups. However, an IP group has a multicast address, therefore, the ARP Protocol cannot find the MAC address corresponding to the packet forwarding on the data link layer. What happens next depends on whether the following data link layer supports physical multicast addresses. Most local networks support physical multicast addresses. Ethernet is one of them. The physical address of the Ethernet is 6 bytes. If the first 25 bits of the Ethernet address are 0000000100000000010111100, it is defined as the physical multicast address of the TCP/IP protocol. The remaining 23 bits can be set to a group, to convert an IP multicast address to an ethernet address, the multicast router extracts the lowest 23 bits from the class d ip address and inserts them into the Ethernet physical address. However, the Group Identifier of the Class D address is 28 BITs, which means that the five bits are not used. This means that 2 ^ 5 (32) IP multicast addresses are mapped to one multicast address, the ing is one-to-one. The host may receive groups that do not belong to the groups it involves. Therefore, the host must check the IP address and discard any groups that do not belong to it.