Finite Buffer Router Architecture

IP routers are modeled on a finite buffer architecture that operates in a slot-based processing mode or a central processing mode. In the slot-based processing mode, the IP module is composed of the following:

  • A routing element
  • One or more interface slot elements
  • An interconnecting backplane

The routing element contains the routing process and memory to hold packets to be routed. Unprocessed packets are stored in a queue, then forwarded to the appropriate slot once processed.

Each interface slot element contains the following:

  • One or more interfaces
  • Memory to hold incoming and outgoing packets (either in a common shared RAM or in physically separate RAMs)
  • A processor for local processing of packets

The backplane interconnector handles the transfer of packets between slots and the routing element.

Figure 213: Slot-based Processing Mode

Slot-based Processing Mode

The central processing mode can be viewed as a degenerate version of the slot-based processing mode — it has no slots and network interfaces feed directly into and out of the routing queue.

Each node intended to use IP has an IP routing module, which contains a dispatcher process that spawns the various routing processes. The three types of routing modules — centralized, slot-based, and cloud — are illustrated below.

Figure 214: Centralized IP Routing Module

Centralized IP Routing Module

Figure 215: Slot-based IP Routing Module

Slot-based IP Routing Module

Figure 216: Cloud IP Routing Module

Cloud IP Routing Module

The following table lists the process models used by the IP model suite.

Table 110: IP Process Models

Process Model Description
ip_dispatch Implements IP routing functions, and fragmentation and reassembly. IP packets arriving on any interface are routed to the appropriate output interface based on their destination address.

The ip_dispatch process requires a fixed amount of time to route each packet. Packets are forwarded on a first-come, first-served basis.
ip_igmp_host Implements the Internet Group Management Protocol (IGMP) in hosts. It is a child process of ip_dispatch.
ip_igmp_rte_grp Implements IGMP in routers. The ip_igmp_rte_intf process spawns a child instance of this process for every multicast group in the subnetwork.
ip_igmp_rte_intf Implements IGMP in routers (with ip_igmp_rte_grp). ip_dispatch spawns an instance of this process for each multicast-enabled interface on a router.
ip_icmp Implements ICMP Echo Request/Response (ping) message generation and handling. This is a child process of ip_dispatch and is created once ip_dispatch configures its interfaces.
ip_pim_sm Implements the Protocol Independent Multicasting - Sparse Mode (PIM-SM) multicast routing protocol in router nodes. This is a child process of ip_dispatch.
ip_output_iface Implements queuing and dropping mechanisms. This is a child process of ip_dispatch. IP packets arriving on any IP output interface are sent to this process if the interface supports a queuing mechanism.

This process model implements the following queuing mechanisms: FIFO (First in, First out), WFQ (Weighted Fair Queuing), Priority Queuing, Custom Queuing, etc.
ip_rte_central_cpu This child process of ip_dispatch implements the centralized processing mode.
ip_rte_distrib_cpu This child process of ip_dispatch implements the routing element in the slot-based processing mode. One instance of the ip_rte_slot process is spawned for each slot with active connections.
ip_rte_slot This child process of ip_rte_distrib_cpu handles all traffic designated for a slot.
oms_basetraf_src Handles background utilization traffic in routers.