Data Connection's Multicast IP Routing products are all part of the following architecture.
This architecture allows the components to be combined in a variety of ways to create routers with specific functionality, and allows seamless integration with Data Connection's suite of unicast IP Routing products through DC-RTM. These IP routing products include DC-ISIS, DC-OSPF, DC-RIP and DC-BGP, which can be used to provide the multicast routing protocols (eg DC-PIM) with the route data to build distribution trees.
Architecture Advantages
Performance and Scale
The product architecture has been designed to provide optimum performance and scalability, in particular:
- distribution of components to take full advantage of multiprocessor systems
- customization of scheduling to optimize CPU usage
- data transfer between processes takes place in the original memory buffers to avoid data copying on the mainline path
- all information on interfaces, neighbors, adjacencies and routes is held in balanced binary trees to enable rapid searching and with no maximum sizes (limited only by memory and CPU constraints)
- incremental updates and route caching are used to minimize CPU expenditure
- data structures have been carefully designed to provide the right trade-off between performance and occupancy on a case-by-case basis.
Distribution and Redundancy
The Multicast IP Routing products are developed within Data Connection's High Availability Framework (HAF). This allows them to operate in a redundant architecture, providing robust fault tolerance and high availability.
The architecture of Data Connection's multicast routing products allows support for a wide variety of distribution and redundancy models.
- Scalable algorithms ensure the products are extremely efficient in a centralized architecture.
- Any combination of DC Multicast components can be distributed across multiple processors or line cards. This can free up a motherboard to perform CPU-intensive tasks.
- Redundant instances of the Multicast Packet Forwarder ensure that the ability to perform multicast packet forwarding is not affected by failure of other parts of the system, allowing five or six nines availability.
Fault Tolerance - Restart
Multicast routing protocols are soft state, which means that multicast routing information is relearned from neighbors at periodic intervals. In most cases, this information can also be explicitly solicited to reduce the latency involved.
Data Connection's multicast routing architecture takes a restart approach to fault tolerance, which is complemented by these features of the protocols.
- Data Connection's High Availability Framework detects the failure of a DC Multicast instance and automatically starts a new instance.
- Product configuration information is injected into the new instance from persistent store.
- Protocol state is re-synchronized with peers on the network.
- The component re-learns state across its interfaces to other components.
Portability and Modularity
Data Connection's Multicast products are fully portable and run independent of the host operating system and software/hardware architectures. For more information on the portability of Data Connection products please see N-BASE page.
The architecture is highly modular, and the presence of well-defined, public interfaces enables any particular component or components to be replaced (making use of appropriate stub code to map to and from the format of the DC Multicast interfaces). Therefore, unless stated explicitly, it should be assumed that any given component in this architecture could either be Data Connection software or independently developed software.
Components
Multicast Routing Protocol Manager
The MRPM component is a multicast routing protocol stack, typically an implementation of a single multicast routing protocol for a single domain, eg DC-PIM-SM.
Multicast Group Membership Discovery (MGMD)
MGMD components handle multicast group membership. Multicast group membership is communicated using DC-IGMP (for IPv4) and MLD (for IPv6). An MGMD component is typically an implementation of a single MGMD protocol, although this architecture does not prohibit an MGMD component from implementing more than one MGMD protocol.
A multicast router may contain multiple MGMD instances to provide scalability and modular support for multiple MGMD protocols. Each MGMD instance owns a subset of the network interfaces on the router.
Unicast Routing Components
The Multicast Routing Information Base (MRIB) is a database of unicast routes to particular destinations that traverse only multicast-capable routers. Information from the MRIB is used by the multicast protocols to set up routing tree state and forwarding rules for multicast traffic. Every multicast router contains exactly one MRIB.
The MRIB is populated by unicast routing protocol components, such as M-ISIS and MBGP components. These protocols are extensions of ISIS and BGP that can handle the existence of different topologies for unicast and multicast routing in the same domain. If there is no difference in topology, a routing protocol such as OSPF can be used.
It is not necessary to populate the MRIB with routes learned from a protocol like M-ISIS. Instead (or in addition), it is possible to configure a static route in the MRIB. In DC-RTM, this is provided by the Static Routes MIB.
The DC Multicast components are architected to allow seamless integration with Data Connection's suite of unicast IP Routing products. The IP routing product family includes DC-ISIS, DC-OSPF, and DC-BGP that can be used to populate the MRIB as described above.
Related links:
For more information about Data Connection's Multicast IP Routing products and expertise contact dciprouting@dataconnection.com.