While multicast technology has been widely deployed in IPTV networks it has hardly found its way onto the public Internet, largely through lack of support by ISPs. This did not matter much when most traffic was one-to-one unicast by nature, but is becoming a handicap with the advent of live HD OTT video services. We do have LTE Multicast for mobile services, but it remains to be seen whether this will gain significant traction among operators beyond major live events when there are many people receiving the same content in a given cell.
But now a new protocol is arriving on the table to enable end to end multicast across unmanaged networks, whether fixed or mobile, that include segments lacking support for any existing multicast protocol. Called Automatic Multicast Tunneling (AMT), this emerging standard tunnels through non-multicast networks using UDP (User Datagram Protocol) to deliver content from core networks right out across the last mile to the user’s home. AMT is not being proposed as an alternative to IP multicast, since it inevitably generates more traffic as a result of the extra signaling involved to traverse unmanaged networks. But it is so far the most viable solution for the Internet since it builds on existing well established IP protocols and will enable automatic configuration.
It does exploit work already done for IP multicast, such as SSM (Source Specific Multicast), which enables delivery of multicast packets from a specific source address requested by the receiver. This reduces unnecessary bandwidth consumed when packets emanate from multiple sources, as well as enhancing security by restricting the flow to a single path through the network. The alternative to SSM is ASM (Any Source Multicast), which enables access to multicast IP packets wherever they are.
AMT is effectively a client/server application, requiring software in the end device communicating with some host system. The client software is now available open source, while the EBU (European Broadcasting Union) Technology and Innovation team is developing an AMT Multicast test bed with practical information and real-life use cases.
AMT is being proposed not just for video distribution, but also software updates and real time data feeds, as well as two way services like collaboration and video conferencing. These applications usually do not benefit from multicasting at present since they go over the Internet. AMT uses UDP encapsulation to tunnel multicast traffic through networks that cannot support it. This enables sites, hosts or applications that do not have native multicast access to a network or source to request and receive SSM and ASM (Any Source Multicast) traffic from a network that does have such capability.
AMT achieves this through the existing Internet Group Management (IGMP) and Multicast Listener Discovery (MLD) protocols. These allow the system to manage, communicate, and execute changes in multicast group membership in real time to ensure that end users receive data when they want but that no unnecessary bandwidth is consumed when they do not. In other words they prune back multicast transmission to network nodes that are upstream of at least one user wanting to receive a multicast. If there is no user downstream of a node it does not receive the relevant multicast packets.
IGMP was developed for IPv4 hosts and routers, enabling them to report their IP multicast group memberships to neighboring multicast routers. One point here is that an IP multicast router may itself belong to one or more multicast groups. In that case it acts both as a multicast router itself in order to collect the membership information needed by its multicast routing protocol and as a group member to inform other neighboring multicast routers of its current memberships. IGMP is also used for other IP multicast management functions.
MLD is the equivalent of IGMP for systems using IPv6, which will over the next decade or more replace IPv4 as the protocol of the Internet, having been developed primarily to increase the address space, as is essential given the proliferation of IP connected devices being driven by the Internet of Things.
Whichever of IGMP or MLD are used, the host or server portion of the protocol runs on a gateway to generate group membership updates that are sent as AMT messages to a relay. This relay in turn runs the router portion of IGMP or MLD protocols to process the group membership updates in these AMT messages and so execute the required changes in multicast forwarding. This is done by sending incoming multicast IP datagrams to gateways according to their current group membership state.
The chief function then of AMT is the handshaking, encapsulation and decapsulation required to transport the IGMP or MLD messages, along with the multicast IP datagrams, between the gateways and relays. These IGMP and MLD messages are themselves replicated and encapsulated as complete IP datagrams within AMT control messages, which are all sent via unicast UDP.
In this way AMT brings several advantages over alternative manual approaches. An obvious one is simplicity by avoiding the overhead of manual provisioning, establishing and maintaining GRE (Generic Routing Encapsulation) tunnels between two locations. The receiving network simply issues AMT advertisements to a desired source using an advertised any-cast prefix. The rest of the tunnel establishment process then happens automatically without any additional configuration.
The second advantage is resiliency, because the gateways automatically find the closest relay and if this is suddenly unavailable or unreachable, the routing table re-converges on the next closest relay. It is also more efficient because AMT allows transit routers on the path to perform flow-based load balancing to share traffic between links.
In the absence of AMT there have been some proprietary efforts to extend multicasting over unicast links, one being nanoCDN from French CDN technology vendor Broadpeak. This involves a light client in a home router, gateway or set top that intercepts an incoming unicast stream and seeks an alternative multicast stream of the same content. Under this scenario, the operator works with a content owner to make the most popular channels such as live sports available multicast. The nanoCDN client then switches to this multicast stream instead and converts it to unicast ABR (adaptive bit rate) in the set top or gateway for onward distribution in the home to multiscreen devices, such as laptops or tablets which have requested it. No software or configuration changes are required in the end device.
CDN provider Akamai has taken a somewhat similar tack for final off-net distribution over the last mile, in this case involving negotiation with ISPs to access their multicast capabilities, which of course might include something like nanoCDN. The idea is that Akamai Edge servers can then stream ABR fragments over UDP and in this case does not require specific software integration on the gateways to make it work on all OTT devices. It does however require the end devices themselves to run an appropriate client stack, unlike the nanoCDN approach.
The underlying trend is that operators, vendors, CDN providers and content owners are acknowledging that multicast will be essential for live OTT services at large scale.