This invention relates generally to a method and apparatus for multicast message transmissions to multiple nodes in a data network, and more specifically, to a method and apparatus for maintaining the list of recipients of a multicast message.
The flow of data in a network is accomplished by transmitting data from one network node to another node until the data arrives at the destination node. Various data protocols are used for transmitting data in the network, including TCP/IP HTTP and UDP. The protocol selected for a data transfer is based on the type of network, the network topology, the data type and other factors.
Two different protocol classes are available for transmitting data through a network: unicast routing protocols and multicast routing protocols. Unicast routing protocols transmit data from a single source node to a single destination node, typically through several intermediate nodes. Although unicast protocols, such as TCP (transmission control protocol), can be used for multicasting, they are not efficient when applied to multicasting as the data is sent multiple times from the source node to each destination node in the multicast group.
Multicast routing protocols transmit data from one or more source nodes to multiple destination nodes, and are designed to be more efficient when sending data to a large number of nodes. The multiple destination nodes are defined as members of a multicast group such that each member of the group receives data addressed to the group. In multipoint-to-multipoint multicast, each receiving node in the group receives data from every sender. If the multicast group is limited to a single sender, this is referred to as point-to-multipoint multicast. Exemplary multicast groups can include the group of all nodes receiving a software update generated at the source node or an update to a node address list that is directed to a network device, such as a switch. Multi-participant videoconferences, distance learning (where a teacher communicates with geographically dispersed students) and multiple emails to a news group are also multicast data transmittals. In one exemplary multicast routing protocol, the sending node transmits data to each multicast group member through a single transmission to a multicast group address. An underlying multicast delivery mechanism delivers the data to all of the destination receivers who have registered to be a part of the multicast group. Known techniques such as multicast data flooding and reverse-path forwarding provide the mechanism by which the data is received by all routers in the network that lie on a path to a receiving node in the multicast group.
In one conventional router implementation, each network router stores a linear list of multicast group members (indexed according to the destination address or destination identifier (DID) of the group members) for those multicast groups in which a destination node connected to the router is a member. In certain applications, the router stores multicast group lists for all possible multicast sessions that might occur on the network, regardless of whether the a particular link from the router is actually participating in a multicast session.
The contents of each destination address entry in the multicast linear list include various nodal and protocol attributes of the destination address, i.e., data transmission parameters. Generally, the contents describe where (over which output port) and when (with what priority) to send the packet to the destination address and the nature of any transformations required in the packet header to ensure the packet reaches the destination address. For example, an entry can include a queue identifier indicating the router output port through which the data must be sent to reach the destination address. Another item in the entry identifies any packet header transformations that are required to satisfy the requirements of the communications medium and the network protocol of the destination address (which may be different than medium and protocol of the source node). A list entry can further include a descriptor of the shortest path from the router to the destination node. At a minimum, information is required for each entry to indicate which headers to prepend to the packet for transmission on each of the outgoing medium. The linear list does not include a link at each entry; rather the entries are arranged in sequence. The last list entry includes an indicator asserting that the end of the list has been reached.
For some applications, namely emulation of broadcast in a virtual local area network environment, a multicast group list is stored in multiple versions, where each stored list includes all members of the group, minus one. Since there is one omitted member from each list, the number of multiple lists for a group is equal to the number of entries in the list. Significant storage capacity is required at each router to store these multiple versions for only a single multicast group.
When a multicast packet arrives at a network router, the sending node and the intended multicast group are determined from the received packet. This information is then used to select the appropriate group list where the source node is the omitted node from the list. Thus the packet is transmitted to the group, but since the source node is not included within the group list, the packet is not transmitted back to the source.
In certain embodiments, the members of the group may be local area networks, each comprising a plurality of nodes, rather than individual network nodes. The source node, which is thus a network, transmits the packet to all nodes within its network and requests transmittal of the packet to a multicast group. In this situation, it is particularly advantageous to prohibit transmittal of the packet back to the source node, since the packet would then be resent to each node of the source network, unnecessarily consuming significant bandwidth within the source network.
Rather than omitting one network node from each group list, according to another multicast technique, each linear list begins with a different member of the group. See
The router classifying engine then reads the address and flag information and selects the correct multicast group list by matching the source address with the first entry in a group list. If the skip first flag is set, the data is not sent to the first address in the group.
If there are N members in a multicast group, each serving as a source for a multicast packet, and the packet must be sent to all other members of the group, then N lists are required. But each list has (N−1) entries because the packet is not sent back to the source node. Thus (N)(N−1)memory locations are required to store all group lists, where each list includes all members of the group except one. Therefore, the storage requirements at each router can be considerable if N is relatively large, as it would typically be.
In another embodiment where the data is transmitted to the receiving nodes of the multicast group in a specific order, then additional group lists are required, with each group list setting forth a different order for the receiving nodes. See
To reduce the memory storage requirements at each network processor, for example, a router, where a copy of the multicast group list is maintained, a multicast group list according to the present invention is linked in a circular structure and allows for any list entry to serve as the starting point for processing through the list. In one embodiment, the chosen starting point for the list is the source node of the multicast data and the data is not transmitted back to the source node. Each entry in the multicast group list is linked to another entry, forming a circularly linked list. The processing engine proceeds through the multicast link list to prepare the multicast packet for transmission to each member of the group. When the addressing mechanism returns to the starting point, the link list has been completely traversed and the process terminates.
The present invention can be more easily understood and the further advantages and uses thereof more readily apparent, when considered in view of the description of the invention and the following figures in which:
Two representative interleaved circular multicast group lists (A and B), according to the present invention, are illustrated in
Each entry in the multicast list according to the present invention contains a pointer or linking information to another entry in the multicast group, and the last entry of the list points back to the entry point. In
The exception to this process is the entry point, A1 in the example above. The entry point is skipped, i.e., the retransmission attributes for transmitting the packet to the entry point address stored there are not processed and a copy of the data packet is not created for the entry point address. Thus the multicast packet will not be transmitted to the entry point address. Instead, only the link information of the entry point is read so that the processing engine immediately proceeds to the next linked entry in the circular list.
There are several different techniques for identifying the entry point to the multicast list. As discussed above, in some applications it is not necessary, nor is it correct, to transmit the data packet back to the source node. Therefore, according to one embodiment of the present invention, the received data packet identifies the entry point and in this embodiment the source node is identified as the entry point. According to one embodiment, the header of the received data packet includes information from which the value of a skip flag is determined. If the skip flag has a predetermined value, then the source address is skipped during processing so that the data packet is not sent back to its source. Alternatively, the received data can identify another address as the entry point and therefore this address will be skipped and not receive the multicast data packet for this session. According to another embodiment, the received data packet indicates that it is a multicast or broadcast packet and the network router, in particular in applications such as a virtual local area network, is programmed to not retransmit the packet back to its source. In yet another embodiment, the entry point for the group list is determined external to the information contained in the data packet. For example, the device may be preprogrammed to always enter a specified multicast list at a predetermined entry point or the entry point may be determined based on the router input port on which the multicast packet was received.
As the process is executing, the first (i.e., the entry node) and the second processed node of the multicast group are saved in the multicast thread for use in determining when the process returns to the entry point. Whenever that node is reached for the second time, the process terminates.
The use of a circular multicast list according to the present invention significantly reduces the memory storage requirements at the network processor, as compared with the prior art. Only a single circularly linked list is required for each multicast group, with the entry point determining the group member to skip or omit from the group for this multicast session. In the prior art, each multicast group list is linearly linked and therefore must be replicated a number of times equal to the number of entries in the list, with each list replication omitting one group member, which does not receive a copy of the multicast packet.
If the starting point is to be skipped, then the process moves to a decision step 28. If the starting point is not skipped, then the process continues to a step 29 where the data transmission parameters for the destination node of the list entry are determined and a copy of the multicast packet is prepared in accordance with those parameters. The process then continues to the decision step 28 to determine whether the list has been completely traversed and has returned to the starting point. If the list has not been traversed, the decision step 28 returns a negative response and the process returns to the step 29 for processing the next destination address. If the decision step 28 returns an affirmative answer, because the next entry is equal to the starting point or to a null next entry pointer then all destination addresses in the list have been processed and the multicast packets are enqueued for transmission at a step 30.
In another example shown in
As indicated in
Destination addresses can be dynamically added and deleted from a multicast group list. Because these additions or deletions may occur while the list is active, i.e., in use to send data packets to the destination addresses, a process that does not interfere with the multicasting of the packets must be utilized. Assume that B5 is to be deleted from the “B” multicast list. First the pointer associated with B4 is modified to point to B6. Then the pointer from B5 to B6 is removed. Each multicast group has an associated thread of status bits, which are cleared whenever an entry in the group is added or deleted. The bits are set active when the group is traversed and thus provide an indicator that the group list has been successfully modified.
Number | Name | Date | Kind |
---|---|---|---|
5095480 | Fenner | Mar 1992 | A |
5295262 | Seigh, II | Mar 1994 | A |
5528588 | Bennett et al. | Jun 1996 | A |
5577246 | Priddy et al. | Nov 1996 | A |
5875189 | Brownhill et al. | Feb 1999 | A |
5924098 | Kluge | Jul 1999 | A |
6014690 | VanDoren et al. | Jan 2000 | A |
6188690 | Holden et al. | Feb 2001 | B1 |
6219352 | Bonomi et al. | Apr 2001 | B1 |
6310875 | Nichols et al. | Oct 2001 | B1 |
6310879 | Zhou et al. | Oct 2001 | B2 |
6321270 | Crawley | Nov 2001 | B1 |
6327621 | Lee et al. | Dec 2001 | B1 |
6396834 | Bonomi et al. | May 2002 | B1 |
6574194 | Sun et al. | Jun 2003 | B1 |
6581063 | Kirkman | Jun 2003 | B1 |
6661789 | Cankaya et al. | Dec 2003 | B1 |
6762995 | Drummond-Murray et al. | Jul 2004 | B1 |
6822958 | Branth et al. | Nov 2004 | B1 |
7075928 | Branth et al. | Jul 2006 | B1 |
7730220 | Hasha et al. | Jun 2010 | B2 |
20010014096 | Zhou et al. | Aug 2001 | A1 |
20020021701 | Lavian et al. | Feb 2002 | A1 |
20030088694 | Patek et al. | May 2003 | A1 |
20120002546 | Sundararaman et al. | Jan 2012 | A1 |
Entry |
---|
Knuth, Donald E. “The Art of Computer Programming”, 2nd Edition, vol. 1 “Fundamental Algorithms”, 1973, pp. 228-231 and 270-273. |
Knuth, Donald E. “The Art of Computer Programming Volume 1/Fundamental Algorithms”. 1997. Addison Wesley Longman. 3rd Edition. pp. 1-45 (254-296). |
Number | Date | Country | |
---|---|---|---|
20030120806 A1 | Jun 2003 | US |