Watch Telephony's Security Webcast Find out what Verizon Business is doing with managed network security and how service providers can cash in with this new service. Watch Telephony's Webcast today.

Telecom carriers are competing with cable companies to deliver new IP video services, variously called IPTV or switched digital video (SDV). Many of these services will require multicasting, or sending a single IP stream to multiple destinations. In IP networks, Internet Group Multicast Protocol (IGMP) is the mechanism that enables multicasting and supports channel-change requests from individual viewers.

IGMP can be implemented in a variety of places in the network, but the location where IGMP fans out from point to multipoint can vary. There have traditionally been tradeoffs with each of two common approaches:

• When IGMP processing is closer to the core of the network, carriers can share the cost of the IGMP processor over a larger number of downstream viewers, but network transport costs are higher.
• When IGMP processing is closer to the edge, the cost of IGMP processing is higher, but transport costs are lower.

This article explains these trade-offs in IGMP implementation, and shows how carriers can eliminate them with new embedded IGMP techniques.

The cost savings of pushing IGMP to the edge arises from lower transport costs—a multicast stream is transported once to an IGMP processor near the edge and then transported over a short distance to multiple viewers. Since moving IGMP out to the edge reduces overall network transport requirements, carriers would ideally like to put the IGMP fan-out point as close to the viewers as possible. However, they have always had to weigh the cost of IGMP processing against the transport savings.

In initial implementations of IPTV, IGMP functions were performed on routers that were centralized deep in the network. The network’s access systems carried multiple copies of the multicast streams from external routers to the set top boxes.

The problem was that the solution wouldn’t scale. As video networks grew larger, it would become impossible to distribute individual streams to every subscriber. Somehow, carriers had to find ways to move IGMP processing closer to the edge.

Fortunately, the cost of IGMP processing has been declining steadily. The first improvement was to integrate IGMP into the access system by developing dedicated IP resources packs. These IP resources packs performed the multicast processing on a dedicated IGMP card, usually at a mid-tier aggregation point.

Integrating IGMP into the access system provided some transport savings, but it created other issues for network planners. The IGMP resources packs often needed to be redundant, so IGMP cards took up two slots in the access system chassis, eliminating slots that would otherwise be used for revenue generating service delivering Ethernet, ADSL2+, or PON. And while the IGMP fan-out point was moved closer to end users, it wasn’t an optimum solution because it wasn’t as close as possible. As a result, IGMP resource packs still forced carriers to transport lots of individual streams to end customers. In addition, these dedicated IGMP resources packs require special rules and introduce new complexities in network engineering and planning dictated by their placement in the network.

The latest solution to this issue takes advantage of further declines in the cost of IGMP processing. It is now possible to push the multicast function all the way to individual service cards in remote terminals. By embedding IGMP in service cards that also deliver services such as Ethernet, PON or ADSL2+, the multicasting function no longer occupies slots that could potentially generate revenue. In addition, there are four other advantages of embedding the multicast function on the service cards:

• It achieves maximum transport savings, since service cards are always located as close to the end customer as possible. (IGMP resources packs or IP routers are typically located farther away from the edge.)
• It scales naturally. As service cards are added, so is IGMP processing power, so each new group of users does not diminish the IGMP processing performance for other users.
• It ensures stable channel-changing performance, because each service card’s IGMP processor serves only 50 or 60 customers, rather than the hundreds or thousands supported by an IP pack or router. This avoids the “half-time at the Super Bowl” problem. By serving only 50 to 60 customers, the difference between normal and peak load is not very large, and channel-changing performance remains stable across all loads.
• It increases security, because a denial of service attack from any one customer’s connection can affect 50 to 60 customers at most, rather than the hundreds or thousands it might affect with IGMP processing in other locations.

IGMP processing has ridden the same price/performance curve as other embedded technologies, and it is now cost effective to place IGMP all the way at the network edge. By doing so, carriers can increase the performance, security, and scalability of their IPTV networks while keeping transport costs to a minimum.

Don McCullough is director of product marketing for Entrisphere.

Visit Entrisphere online.