Concepts related to multicast

Class D addresses are used for multicast, which starts with 1110, and the remaining 28 are used to identify multicast addresses (the remaining 28 are non-structured addresses ), MAC addresses starting with 01005e are used to represent MAC multicast addresses, while the remaining 23 are used to identify multicast. IP multicast addresses have a ing relationship with MAC addresses, that is, the latter 23 of IP multicast corresponds to the latter 23 digits of Mac multicast, so that multiple IP addresses may correspond to one MAC, and Arp is used for resolution between them.

If a LAN uses multicast, all hosts must be added to an all-host group (the multicast address is 224.0.0.1), but the Host can decide whether to accept multicast. Before a multicast traffic decides to be transmitted, the router needs to know that the Host wants to receive multicast, IGMPv1, and IGMPv2, to find whether the multicast group members are in the directly connected subnet, IGMP uses a querying device to request and report Host joining and leaving multicast groups. It has two types of messages: query messages used by multicast routers to discover Member, another is report message, which is sent by the host, which reports that the host wants to join the multicast group, and IGMP continuously sends the query message to 224.0.0.1 (used to identify all hosts ), one of the hosts in the direct connection CIDR Block sends a report message to report who wants to receive the multicast. However, if the Host wants to join the multicast group, it does not have to wait for the query message, host can actively send a report message to 224.0.0.2 (identify all multicast routers in the multicast group)

IGMPv2 provides leave meassage. The host sends this exit message to 224.0.0.2, indicating that the host wants to leave the multicast group. The query router sends a query message (specific-group) information from the port that receives and leaves the group. If other hosts want to join the multicast group, send a report message of specific-group to the query device, if the host does not respond to this group-specific query message within a certain period of time, it indicates that no local member is in this group.

In vswitch forwarding multicast, because the IP Group is broadcast to a Layer 2 multicast address, multicast must be sent to all vswitch ports. When a host reports member information to the multicast router, multicast is sent from the vro to the vswitch. Because multicast uses multicast addresses to send traffic, it does not know the actual target mac address, so it sends it to all the vswitch ports, VLAN division can be used to solve this problem, but VLAN cannot be used to dynamically add or delete members. In this way, the switch must process each multicast packet, increasing the processing latency, the performance of the vswitch is reduced, so a router to switch solution-CGMP, CGM allows the vswitch to learn information about the group members from the multicast router. In this environment, rotuer is a CGMP server, and switch is a Client. When the router receives a multicast packet, it immediately creates a CGMP packet, which is sent to a well-know address, all switches can receive this package. Then the switch explains this package and creates an f Orwarding table.

In each physical segment, a specified vro is selected, indicating that the vro constructs a distribution tree and connects all the members of a multicast group to ensure efficient transmission. This router can copy all incoming packets and send them to its branch tree. Because multicast groups are dynamic, the distribution tree must also be dynamically upgraded, adding a branch to a distribution tree is like adding a new member to a multicast group. On the contrary, if the branch has no receiver, the Branch has just been deleted.

Two distribution trees can be used in multicast routing: 1. source specific 2. shared: Specifies the source method to process the Spanning tree for the source of each multicast group. For example, if 10 members are in 10 isolated subnets, create 10 different routing trees based on 10 Multicast Groups, and specify the source to use the shortest path from the source to the destination to minimize the latency, the source-based distribution tree uses a Reverse Path Forwarding RPF mechanism. When a router receives a source multicast packet, it forwards the packet on all its ports except the port that receives the packet. However, the forwarding only takes place on the link that provides the shortest path to return to the sender, if the package reaches a non-Shortest Path, the package is discard.

A parent link is provided to indicate that the link between the router and the source is called a parent link, and a port is called a child link at the router.

In the shared-tree mode, all packets are sent to the multicast group along the distribution tree, regardless of the source to be sent. This method reduces the processing time, however, this results in large end-to-end latency.

However, unlike the spanning-tree mechanism, different Multicast Groups define different distribution trees. If a device wants to receive data, it must join the shared tree of the group, the Multicast Routing Protocol detects that an on-demand route can be reached to establish a distribution tree.

Multicast packets use the TTL domain of the IP Header to limit their accessible range. Each time a router passes through, the TTL value is reduced by 1. If the TTL expires, the packet is discarded, if the TTL of Packet is greater than the TTL threshold of the API, it is forwarded. If the TTL is smaller than the API threshold, it is discarded.

0 is restricted to the same host and never sent to any interface

1. It is restricted to the same subnet and never forwarded by the router.

15 restricted to the same site, organization, or department

63 restricted in the same region

127 worldwide

191 workwide, limited bandwidth

255 unrestricted in scope; global

The IP multicast routing protocol is used to discover multicast groups and create a distribution tree for each multicast group.

Client to router: IGMP

Router to Switch: CGMP

Router to Router: DVMRP, PIM, MOSPF, CBT

The multicast routing protocol can be either of the following two methods: dense-mode routing and sparse-mode routing)

How to choose these two methods depends on the distribution of multicast group members throughout the network, if almost all routers in the network distribute multicast information for each multicast group, use Dense-mode. To maintain the distribution tree, the Dense-mode multicast routing protocol uses the intermittent flood Network Multicast information. The Dense-mode applies to group members who are densely distributed across the network and have sufficient bandwidth to tolerate flood.

The sparse-mode routing protocol is used for each Multicast with only a few vrouters (not that each multicast group has only a few Members). It means that multicast group members are widely dispersed, for example, Internet Muticast and sparse-mode assume that the network bandwidth is limited and sparse-mode does not use flood. At the beginning, it creates an empty distribution tree, only when a member requests to join a multicast group, it adds a branch to the distribution tree.

The dense-mode Routing protocols include Distance Vector Multicast Routing Protocol (DVMRP), Multicast Open Shortest Path First (MOSPF), and Protocol Independent Multicast Dense Mode (PIMDM)

Most DVMRP is used for the multicast trunk (MBONE) router, which uses the reverse path flood (reverse path flooding). When DVMRP receives a packet, it flood this package on all the paths it connects, except the receiving path, so that this package can reach all the LAN, if a network segment does not have any multicast group members, then the router sends a cut message to return the distribution tree. This person cut the information to prevent later packets from being sent to this region without any members, DVMRP uses its own integrated routing protocol to determine the path of the packet return source. This on-demand routing protocol is similar to RIP. It is based on hop counts. In order to allow new hosts to join multicast groups, DVMRP intermittent flood and DVMRP are rarely used in large networks. DVMRP has poor scalability because it relies on Flood.

MOSPF (not supported by Cisco) depends on its Integrated OSPF. MOSPF is applicable to individual routing domains. For example, a network is controlled by a separate organization, and OSPF is a link status routing protocol, MOSPF adds the multicast information to the OSPF link status advertisement. In an OSPF/MOSPF network, each router maintains a detailed network extension based on the link status information, a mospf router uses link status advertisements to learn which multicast group is activated in the connected LAN. It constructs a distribution tree using this information. MOSPF forwards packets based on the packet source and destination address, A separation Shortest Path distribution tree based on each source-group is established. The distribution tree is re-computed when the network extension changes and the cache expires, MOSPF is suitable for activating a small number of soure-groups at the same time. It is not recommended that MOSPF be used in unstable environments.

Similar to DVMRP, pim dm uses the opposite path flooding (reverse path flooding). When pim dm receives a package, it flood the package on all the paths it connects, except for the receiving path, if a CIDR block does not have any multicast group members, the router sends a cut message and returns the distribution tree. The protocol is independent, meaning that it does not depend on any specified on-demand routing protocol, this principle applies to dense-mode and sparse-mode. PIM can use all the on-demand routing protocols. PIM applies to the very close distance between the sender and the receiver, and also to very few senders and many receivers, and high traffic.

Two multicast routing protocols for Sparse-mode:

Protocol Independent Multicast Sparse Mode (pim sm) and Core-Based Trees (CBT)

Pim sm applies to scenarios where only a small number of receivers and traffic is not frequent. This protocol can process several multicast data streams at the same time and is suitable for WAN or Internet applications. It defines a set point (rendezvous point ), A sender must send data to this set point. Before receiving data, the receiver must register at the set point. The router automatically optimizes the path, PIM can use dense-mode in some multicast groups and sparse-mode in other groups.

In the CBT environment, all Group members share a separate tree, and multicast streams are transmitted on the same distribution tree. The source is not considered. The CBT and Spanning-tree are similar, in addition to creating a separate tree for each multicast group, a Core-based tree can use a separate router or a group of routers as the Core. The router sends a message to join the Core, the core sends a confirmation message to the vro. The addition information does not need to be confirmed by the core. This vro becomes a branch of the distribution tree.

Example:

A client sends an IGMP message, and the next hop router receives the message, records the IGMP source MAC address, and generates a CGMP packet sent to the switch, the switch uses this CGMP information to dynamically create a table item in switch talbe. This table item is the ing between the actual multicast Host address and the switch port.

Article entry: csh responsible editor: admin