12.13.06

Why multicast is irrelevant to the Internet

Posted in Essays, multicast/anycast at 9:14 pm by Sailesh Kumar

Any communication process can be divided into one of the three main categories, unicast, multicast, and broadcast communication. In unicast, messages are sent from one source to one destination; the message may traverse through several intermediate nodes. In broadcast, messages are sent from one source to all destinations which have physical connectivity to the source. Multicast is the process of sending the message from a single source to a subset (containing two or more hosts) of all physically connected destinations. When messages have to be sent from one source to multiple destinations, multicast is generally a much more efficient mode of communication than simply unicasting the message to all destinations. Multicast is efficient because the message need not be replicated for every destination; replication only occurs when the message gets close to the destination. Thus, the links which are close to the source do not require large bandwidth, which is otherwise needed if the same message is unicasted multiple times. Clearly, multicasting has several applications, some of which are multi party video conferencing, broadcasting information to a large number of users, real-time video distribution, etc.

Current Internet supports both unicast and multicast capability however unicast is the most prevalent mode of communication, with multicast being almost non-existent. While many researchers have advocated a wide deployment and use of multicasting in the Internet, a large number of researchers along with the Internet Service Providers (ISPs) have largely remained skeptical about its real benefits. While researchers have been mostly concerned about the added complexity that multicasting will introduce in the core Internet, ISPs are concerned more with the current payment model of the Internet. Presently, ISPs charge users based upon the amount of link bandwidth they use; in such a model, allowing multicasting may reduce the bandwidth usage by individual users, thus the revenues of ISPs may decline. Clearly, in order for the ISPs to have enough motivation to deploy multicasting capability, the payment and service model of the Internet usage needs to be changed.

There are many other complicating factors, which generally dilutes the benefits of multicasting in the current Internet. For instance, it is generally accepted that a large fraction of the total Internet bandwidth is consumed in one-to-one communication. Thus, if only 1% of all Internet communication operates in a one-to-many mode, then there will be little incentive in deploying the multicasting capability. The situation becomes even worse, because the one-to-many mode of communication is generally limited within small local networks, for instance sending emails to all users in a company, etc. There are very few instances where one-to-many communication occurs with destinations distributed across the world. On can argue that the situation is similar to the chicken and egg problem; since ISPs do not support multicasting, there are few services which needs them, and since there are few services which needs them, ISPs do not have enough motivation to implement multicast.

With the introduction of the Overlay networks, the situation has changed considerably. Now, there are many applications and services, which use application layer multicasting over an Overlay network. These services, usually builds a multicast tree with the source at the root of the tree and the destinations at the leaves. The intermediate nodes which are essentially the overlay nodes are arranged in such a way that the overall bandwidth usage in minimized. With the introduction of these Overlay based applications, multicasting has received a new push, and now many argue that ISPs should also deploy the multicasting capability in their network.

While these Overlay networks have demonstrated that multicasting can be useful for a wide variety of services, there are yet many technical barriers. First and foremost, the complication arises due to the presence of hundreds of independent ISPs in the current Internet, and due their conflicting interests. Each ISP wants to provide the best possible service to its own customers, and generally they do not have enough motivation to improve the service received by its non-customers. In such a scenario, an efficient multicast routing which requires construction of multicast trees from source to all destinations can become problematic. Moreover, if a small number of ISPs do not allow multicasting, then it may become problematic to implement multicasting capability by other ISPs. Another problem is that IPSs usually do not advertise the topology of their network to other ISPs, and especially to all sources; without such information, it may become difficult to construct efficient multicast trees.

With these technical barriers, there are other issues too, which generally arises due to the dynamics of today’s Internet usage. Today, a large number of Internet content providers use content distribution networks (CDNs) to distribute their content. These CDNs replicate data at several servers, which are generally distributed across the entire world. With such CDNs, the benefits of multicasting become questionable. When a set of users accesses certain content, then the content for each of the destination may arrive from different sources, thus the communication does not occur in multicast mode, even though the destinations are receiving the same content. CDNs are already extremely popular, and they are likely to become more popular in the near future, thus multicast mode of communication may soon become less relevant.

Another trend that does not bode well with multicast is that, many newly emerging services are less likely to require multicast communication. For instance, video on demand (VoD) has traditionally been hailed as one of the prime beneficiary of multicasting, however, with a wide array of contents available, and freedom to select any content at any point in time, the likelihood that a large set of users are accessing the same content at precisely the same time reduces considerably. Thus, most of the communication will occur in a unicast mode. Video conferencing is another application of multicasting, however in a video conferencing it is unlikely that more than a few users will be participating simultaneously. The only application which will really benefit from multicasting is distribution of live information, events, news, etc. Such events are likely to be watched by thousands of users, and all at the same time, so multicasting can distribute the content much more efficiently. In fact, unicasting may become impossible in such scenarios.

It clearly appears that multicasting is required for such services like the live content distribution. However, traditionally, Internet has not been the popular medium for such services. Television remains the most popular and efficient medium for such broadcasts, and it is likely that i) people will prefer television in the foreseeable future, and ii) content providers will stick to the current transmission technologies. Current transmission methods are much more efficient and are capable of handling millions of users simultaneously therefore a migration to the Internet for distribution of such content is unlikely. In my opinion, the providers of such media will also resist from moving to the Internet.

With so many barriers and problems at hand, i) resistance by the Internet Service Providers (ISPs), ii) resistance by the content providers, iii) natural shift away from the one-to-many communication mode (due to the introduction of CDNs), and iv) relatively few services for which multicasting is essential, it is unlikely that multicasting will become a popular mode of communication in the Internet. Multicasting will remain an interesting topic of study by researchers, because it involves several interesting technical challenges however its deployment in the near future is highly unlikely.

Permalink