All About...Multicast Wide Area Network Services

Overview

Broadcast and multicast (a subset of broadcast) transport services have been available in local area networks (LANs) since their inception, due to the inherent broadcast nature of the network topology. Wide area networks, however, do not have an inherent broadcast topology, and are only just recently offering multicast services.

This document is intended to provide you with the basic information you need to understand wide area multicast network transport services, and how they could possibly benefit you.

Q. What are broadcast and multicast data transmission services?
A. Broadcast transmission is the sending of information from one to all recipients; multicast is the sending of information from one to many, but not all, a subset of broadcast. The term broadcast is familiar to us all as it is the term used to designate radio or television transmissions over the airwaves. Radio and television broadcasts are sent one way to everyone with a radio or television set used to receive the information.

Q. Why are broadcast and multicast data transmission services useful?
A. There are two broad categories of applications that are fueling interest in multicast data services over wide area data networks. The first is for real time transmission of information that is put into the form of data, but is not computer data in the classical sense. A good example is desktop video conferencing, where specialized video conferencing equipment is not needed, but rather all participants use their desktop PC and windowing to see the participants. Data networks are also being proposed as alternatives to traditional analog transmission of video. This is being made possible by advanced video compression techniques. A prime example of this type was the recent transmission of a Rolling Stones concert over the Mbone, the multicast backbone of the Internet.

Q. What is the second broad category of application?
A. The second broad category of application is the extension of file transfer to be one to many rather than simply point to point. Multicast file transfer applications are useful for information service providers, who desire to deliver information electronically over a wide area network to their customers. It also is a very useful way to distribute software updates to equipment that is dispersed over a large geographic network, and as a general method for database update when there is a large number of geographically distributed offices of an organization such as with retail stores or restaurant chains.

Q. How are multicast wide area networks created?
A. There are two principal ways to create multicast wide area networks today; using multicast frame relay or SMDS and a bridged network, and using multicast IP with a routed network.

Q. How do they work and how do they differ?
A. First, let us describe how multicast frame relay services work. The Frame Relay Forum in October, 1994 released the Multicast Implementation Agreement. This agreement describes three multicast configurations: "One-Way Multicast", "Two-Way Multicast", and "N-Way Multicast". These configurations are shown in Figures 1a, 1b, and 1c.

Figure 1a: One-Way Multicast
One-way multicast, as shown in Figure 1a, has a unidirectional multicast Data Link Connection Identifier (DLCI, frame relay address for a permanent virtual circuit, PVC). This PVC connects the source of information ("A" in the diagram) of the multicast group to a "multicast server" in the network. The multicast server then distributes the data to the members of the multicast group (B, C, and D in the diagram) over their own individual DLCIs t, u, and v. The multicast server is generally located at a convenient location by the network provider to minimize transmission of duplicate data in the network. Additionally, the source A has point to point (unicast) PVCs represented by DLCIs b, c, and d between itself and the other members of the multicast group, B, C, and D. These PVCs are full duplex, i.e. data traffic can flow in both directions. In contrast, the multicast DLCI is unidirectional, from source A only.

Figure 1b: Two-Way Multicast
Two-way multicast is shown in Figure 1b. In this case, the source A communicates to the multicast server with bi-directional multicast DLCI a; data from the source is then forwarded to destinations B, C, and D via unicast DLCIs b, c, and d respectively. Responses from B, C, and D may be sent to source A via the multicast server and are delivered to the source over the bi-directional multicast DLCI a. Destinations B, C, and D cannot communicate with each other over this configuration, however, nor to other possible unicast destinations outside of the multicast group. Similarly, A cannot participate in unicast transmissions.

Two-way multicast is an ideal configuration to migrate old IBM based SDLC multidrop lines into a frame relay environment.

A third configuration is N-Way Multicast as shown in Figure 1c. With N-Way multicast, any member can be either a source or a recipient of data. A multicast server (or servers) are present in the network and all PVCs provide multicast operation with multicast DLCIs.

N-Way multicast is useful for teleconferencing and for update of routing tables for routers.
Figure 1c: N-Way Multicast
Thus, most multicast data transmission applications would use the one-way multicast configuration, as it provides flexibility for unicast transmissions as well. Most data network configurations demand the flexibility to be able to communicate on a unicast basis to other points in the network without restriction.
Q. Often, sources and destinations of data reside on a LAN at their local site. How does this work with wide area frame relay multicast services, in particular the one-way configuration that is preferred for data transmission?

A. Frame relay can be the underlying wide area link layer protocol for bridge and router networks. The one-way frame relay multicast configuration will often be used with bridged networks as shown in Figure 2. In this case, a broadcast or multicast LAN MAC address is mapped to the multicast frame relay DLCI, and unicast MAC addresses are mapped to unicast DLCIs and forwarded in an identical manner as with conventional bridges.

Bridging encapsulates the LAN link layer into frame relay as shown in Figure 3, which adds overhead.

Figure 2: Bridging with Frame Relay
Bridging address filters should be put into place to make sure that multicast frames are not sent out on unicast DLCIs in the bridge learning process, and conversely so than unicast frames are not sent out onto the multicast DLCI. Additionally, it may be advantageous to manually assign the broadcast or multicast MAC address to the multicast frame relay DLCI in the bridge's forwarding table.

Figure 3: Encapsulation of MAC Frame in FR (Bridging)
Q. Are there any disadvantages to using frame relay multicast for the wide area multicasting network topology?

A. Multicast frame relay is a relatively static configuration. Network configurations are difficult to change once initially set. This means that it is not feasible to have dynamic multicast groups that can be set up and torn down on demand, based on changing the network configuration.

Q. What about multicast SMDS? How does it compare with multicast frame relay?
A. SMDS is more suited to multicast WAN applications than is frame relay. The reason is that SMDS is a connectionless service, i.e. all nodes on the network are inherently connected to each other as is the case on a LAN. SMDS has multicast and broadcast address space just as LANs do and can be used for multicasting in the same way as nodes on a LAN may use multicast services.

This means that the equivalent of specific PVCs and their DLCIs and CIRs in frame relay do not need to be configured in SMDS; connectionless networks do not have PVCs.

The typical SMDS bridged network shown in Figure 4 is similar in concept to the frame relay bridged network shown in Figure 2. Q. Can the source and destination nodes be directly attached to the wide area SMDS network, thus eliminating the overhead associated with bridging?

A. Yes, there is now at least one vendor (MultiAccess Systems) of SMDS cards that fit inside a PC that supports NDIS, a common API used by LAN network interface cards. This means that the protocol software used, e.g. TCP/IP, can be used with no change from a LAN configuration. This is also true with frame relay, as there are a few vendors of cards fitting into a PC that provide direct connection to frame relay networks. However, as mentioned above, the frame relay network must be explicitly configured by the network service provider to provide multicast services, which is not required in SMDS.
Figure 4: Bridging with SMDS
The network configuration for direct connection would look like that shown in Figure 5. In this case, the source of the multicast information is shown directly connected to the SMDS network using an internal SMDS PC card. In Figure 5, the same computer with the SMDS card is also shown attached to a local LAN, which requires a separate LAN network interface card.

Figure 5: Direct Attachment to SMDS Network Q. Is a multicast IP router network a better network configuration for multicast networks?

A. Multicast IP is likely to become the preferred multicast network configuration. The reason is because multicast groups can be dynamically set up and torn down, providing much more flexibility than with static configurations using multicast frame relay. Q. How are multicast IP networks set up and how do they work?

A. Multicast IP networks are network layer (layer 3) router networks with routing, both for point to point transmissions and for multicast transmissions. However, multicast is today only supported in the TCP/IP protocol stack, which means that TCP/IP must be used as the backbone protocol. This is not much of a limitation, however, as TCP/IP is becoming pervasive, driven by the growth of the Internet, which is TCP/IP based.

To understand how multicast IP works, we must first understand some of the fundamentals of TCP/IP and routing. Routing occurs using network layer IP 32 bit addresses. IP addresses come in four categories; Class A, Class B, Class C, and Class D. There is also a fifth designation, Class E, that is currently not assigned.

Class A, B, and C addresses are all used for point to point (unicast) communications. These addresses are broken into network and host components, with Class A consisting of many hosts and few networks, Class C, many networks and few hosts, and Class B intermediate in allocation between hosts and network IDs. Class D multicast addresses do not have this network and host breakdown but rather consist of one overall address. This is shown in Figure 6.

Routers utilize routing algorithms to determine the optimal route through the network from source to destination. Routers on the market today do not all use the same routing algorithm, although the interior routing algorithms used within an "autonomous system" are usually made to be the same for best performance. An autonomous system is defined as a network controlled by a single administrative authority. For example, the Internet is actually a network of networks, and each network subsystem is typically an autonomous system.

There are also exterior routing algorithms that are used to route data between autonomous systems.

Figure 6: IP Address Types
Unlike unicast addresses that are assigned as relatively permanent addresses, Class D multicast addresses may be assigned to be either relatively permanent or dynamic as temporary assignments. This latter is useful as it provides the ability to create and tear down multicast groups on the fly.

Q. Are there any changes required to the TCP/IP protocol stack to accommodate multicast besides the usage of Class D addresses?

A. Yes, hosts (computers) that participate in multicast groups must also support RFC 1112, an Internet standard. RFC 1112 specifies three levels of conformance; Level 0, with no support for multicast, Level 1, which provides support for sending but not receiving multicast IP data, and Level 2, which provides full support for IP multicasting.

Sources of data that desire to send data to multicast groups but do not have the need to ever be in a multicast group only need to support Level 1. Level 1 multicast service is a small extension; the host only needs to support the sending of data using Class D addresses as the destination address.

Level 2 support requires the ability to be a member of a multicast group and accept multicast data as a Class D IP host. Level 2 hosts also need to provide support for the Internet Group Management Protocol, IGMP. IGMP provides the means for members of the multicast group (i.e., Level 2) to notify the nearest router that supports multicast routing of its presence in the group. IGMP messages consist of queries from routers to local nets, inquiring if there are any multicast group members on that net. Hosts (computers) that are members of a group respond with reports to the router, notifying it of their membership in a group. This dialog is shown in Figure 7.
Figure 7: IGMP Message Dialog
Figure 7 shows the nearest router supporting multicast routing sending out an IGMP query to the local net to which it is attached. Host C on the local net is a member of a multicast group, so it responds with a report to the router. IGMP queries are sent using the IP all-hosts group address (address 224.0.0.1) as the destination address. IGMP reports are sent with the IP destination address equal to the multicast group address being reported, in the diagram Host C's Class D address. If a host on the local net hears a report issued for a group to which it belongs, it does not need to send a report as well and does not do so. When a host joins a new group, it immediately sends a report rather than waiting for a query. Queries are typically sent at intervals of no more than once a minute to keep overhead low. However, when a multicast router starts up, it may issue several closely spaced queries in order to build up its knowledge of local memberships quickly. Queries act as a "keep alive". When no more reports are issued in response to a query, the multicast router knows that there are no local group members, so adjusts it's router tables accordingly.

Q. How are multicast groups created?
A. There are two levels of creation; the actual multicast group, and the members of a particular group that has already been formed. Class D IP group addresses may be semi-permanently assigned similar to today's administration of Class A, Class B, and Class C unicast IP addresses. However, members of the group may decide to join or leave the group dynamically either under their own initiative or by external direction. Additionally, Class D IP addresses may be assigned by a server on a temporary basis; when the group is no longer needed, the Class D address is returned to the server for reuse later.

The dynamic joining and leaving of an established group is analogous to viewers or listeners tuning in to particular TV or radio stations for a period and then turning the station off. The viewer or listener temporarily becomes the member of the group viewing or listening to a permanent station (group). This concept can be useful in data networks as well, as users join a group to temporarily subscribe to an information service.

The dynamic creation and tearing down of groups finds usefulness in both real time and file transfer applications. A video conference using multicast IP Class D addresses is a temporary group by nature, and is set up for the time of the session and torn down after it is over. Similarly, an information service provider can create a new group for the purpose of a promotion of a particular information service, which then is torn down when the promotion is over.

Q. How long does it take to join and leave groups?
A. Members joining a group using the IGMP protocol to inform the nearest multicast router of their presence can do this in seconds. Leaving a group takes longer, however, as the router learns that a host is no longer in the group by not receiving an IGMP Report in response to an IGMP Query. Since Queries may only be sent out at intervals of minutes, it requires that amount of time to learn that a host is no longer in the group. Also, additional time is needed for the routers in the network to update their router tables after status changes.

Q. Can a particular host be a member of multiple groups simultaneously?

A. Yes. There is no reason why a particular host could not be a member of several groups. In Figure 8, the shaded PC is a member of both groups A and B. That means that it can receive information from both groups, either at different times or simultaneously. In the diagram, a server, which is not a member of either group, can be the source of information. However, group members may also be the source of information for their own group or other groups as well as the recipient of information for the group in which it is a member.

If simultaneous group transmissions to a single host who is a member of multiple groups is contemplated or desired, that host needs to have sufficient resources and the network needs to have sufficient bandwidth to handle both transmissions at the same time.

Figure 8: Host in Two Multicast Groups
Q. I am convinced that multicast capability with wide area network service would be useful to my organization. Can I get this service from a carrier, or do I have to create a private network to gain multicast services?

A. Some carriers are offering multicast frame relay services at the time of this writing in mid-1995. Some examples are MFS Datanet and U. S. West. Any of the carriers offering SMDS services can provide multicast service. In the U.S., Bell Atlantic, Bell South, U.S. West, PacTel, and MCI all offer SMDS services. A number of European carriers also are offering or planning to offer SMDS. Also, some carriers today offer router data network services, and can be expected to offer multicast IP services as well in the near future.

Q. What does StarBurst Software do? Can StarBurst help me with my multicast needs?
A. StarBurst Software has introduced the first multicast file transfer protocol product on the market called StarBurst MFTP. StarBurst MFTP operates on top of UDP in the TCP/IP stack and together with multicast wide area network services, based on multicast frame relay, multicast SMDS, or multicast IP, provides a complete solution for the user who desires to send information from one to many reliably over a data network. Contact StarBurst Software for more information.

Glossary of Terms API: Application Programming Interface; the generalized term for a defined software interface for software applications.

Autonomous System: A network controlled by a single administrative authority; routing domain.

Broadcast: The sending of information from one to all in a network.

Class A: A type of unicast IP address that segments the address space into many network addresses and few host addresses.

Class B: A type of unicast IP address that segments the address space into a medium number of network and host addresses.

Class C: A type of unicast IP address that segments the address space into many host addresses and few network addresses.

Class D: Multicast IP group addresses.

Connectionless: Term used to describe data transfer without the existence of a virtual circuit.

CRC: Cyclic Redundancy Check; a mechanism to detect errors in frames.

DLCI: Data Link Connection Identifier, the link address used in frame relay. DLCIs may be unicast or multicast.

Ethernet: An industry LAN standard sponsored by DEC, Xerox, and Intel in the early 80s. Became the basis for the official IEEE 802.3 LAN standard.

Frame: The link layer data entity; data is packaged in frames, for the purpose of transmission over a network. Frames are bounded by flag characters or some other delimiter.

Host: The generalized term for any device that can be a source or sink of information on a network. Generally, a host is a computer.

IETF: Internet Engineering Task Force; the body associated with the Internet that recommends "standards" for use on the Internet.

IGMP: Internet Group Management Protocol, the protocol with which hosts communicate with the nearest router supporting multicast to notify them about membership in a multicast group.

IP: Internet Protocol; the network layer (layer 3) of TCP/IP. Network layer addresses are used by routers for routing purposes. LLC: Logical Link Control; the link layer specified for the IEEE 802.x series of LAN standards. LLC1 is connectionless, LLC2 is connection oriented, and LLC3 is connectionless with acknowledgment. MAC: Med

ia Access Control; the protocol used in a LAN or other shared transmission media for gaining access to the media. Multicast: The sending of information from one to many, but not all members of a network.

Multicast Group: A group set up to receive messages from a source. These groups can be set up based on frame relay or IP in the TCP/IP protocol suite, as well as in other networks.

OSI Model: A layered model of data communications protocols standardized by the International Standards Organization (ISO).

PVC: Permanent Virtual Circuit; a permanent logical connection set up with packet data networks such as frame relay.

RFC: Request for Comment; the document which the IETF uses to create standards for use in the Internet.

SMDS: Switched Multimegabit Data Service. High-speed, connectionless, packet-switched, Titograd-based WAN networking technology.

SVC: Switched Virtual Circuit; a switched logical connection set up on a temporary basis with packet data networks such as frame relay.

TCP/IP: The protocol suite used in the Internet. Rapidly becoming the most important protocol suite used in networking.

Unicast: The sending of information from one to one in a network; point to point.
New Softwares
Platforms and availability
AutoSync enables network managers to schedule simultaneous synchronization of content such as video, audio
StarBurst customers include automotive companies such as General Motors, Ford, Chrysler and Honda;
Newsletter Signup
Contact : 01-234-567800
Email :