Why Two ATM Modes?

Why Two ATM Modes?

Two modes exist for geographic scalability and efficiency. LANE emulates a local-area network and creates many virtual circuits to support each LAN Emulation Client (LEC) in each Emulated LAN (ELAN). Each LEC maintains four virtual circuits to retain ELAN membership. Two are point-to-point connections (one to the LAN Emulation Server [LES] and one to the broadcast and unknown server [BUS]), and two are point-to-multipoint connections (one from the LES and one from the BUS). The LEC connects as a leaf for the point-to-multipoint connection, with the LES and BUS as roots. An ELAN with 100 LECs then has 200 point-to-point connections, and 200 leaf node connections just for the control virtual circuits to maintain VLAN membership.

This scenario does not even count the data direct connections required for data transfers between LECs. Clearly, many more connections can be present in the ELAN to support peer-to-peer communications.

If LECs for an ELAN are geographically dispersed in a campus or enterprise network, the connections can consume a lot of virtual circuits in many ATM switches. And, if the user leases ATM services, they can be billed for each circuit established in the system. From a technical point of view, there are no limitations on the LEC distribution. But practically, LECs within an ELAN should reside in geographically limited areas to reduce the number of virtual circuit resources consumed in the ATM switches throughout the network.

Further, consider what happens when a LEC in one ELAN needs to communicate with a LEC in another ELAN. The traffic must pass through a router because each ELAN defines a unique broadcast domain. ELANs interconnect through routers, just like VLANs.

Frames traveling from a source LEC in one ELAN to a destination LEC in a second ELAN must traverse routers, as discussed in Chapter 9. If a frame passes through multiple routers to get from the source LEC to the destination LEC, multiple data direct circuits must be established so routers can pass the frame from one ELAN to another. A source LEC must establish a data direct circuit to its default gateway, another data direct must be established from that router to the next router, and so on to the final router. The last router must establish a data direct to the destination LEC. Consider the network in Figure 10-1. When an Ethernet frame from VLAN 1 (destined for VLAN 2) hits the switch, the switch LEC segments the frame into cells and passes it to the default router LEC (Router 1). The default router LEC reassembles the cells into a frame, and routes it to the LEC on the next ELAN. This LEC segments the data and forwards the frame (cells) to the next hop router LEC (Router 2) over ATM. When the cells hit Router 2's ATM interface, Router 2 reassembles the cells, routes the frame, and then segments the frame before it can pass the frame to Router 3. Router 3 reassembles the cells, routes the frame, segments the frame into cells, and forwards them to the destination Catalyst. This Catalyst reassembles the cells into a frame and passes the frame onto the destination Ethernet segment.

Each hop through a router introduces additional latency and consumes routing resources within each router. Some of the latency stems from the segmentation/reassembly process. Another latency factor includes the route processing time to determine the next hop. This element can be less significant in routers that do hardware routing (as opposed to legacy software-based routers).

Figure 10-1. A Multi-ELAN Network Data Flow

The hop-by-hop approach was necessary when networks interconnected with shared media systems such as Ethernet. Physical connections force frames to pass through a router whenever a device in one network wants to connect to a device in another network. LANE maintains this model. It honors the rule that devices in different networks must interconnect through a router. MPOA, however, creates a virtual circuit directly between two devices residing in different ELANs.

MPOA, in fact, bypasses intermediate routers but gives the appearance that traffic from a device in one ELAN passes through routers to reach the destination device in another ELAN.

To get from one VLAN to another over ATM can require communication across several ELANs. The following sections compare how traffic flows within an ELAN as opposed to when it needs to cross more than one ELAN.

Intra-Subnet Communications

When Catalysts need to communicate with each other within the same ELAN, the Catalysts use LANE. Intra-subnet (or intra-ELAN) communications occur whenever devices in the same broadcast domain trunk to each other. The Catalyst LANE module was exclusively designed to support LANE operations with high performance to handle flows at LAN rates.

Details for intra-subnet communications are described in Chapter 9.

Inter-Subnet Communications

As discussed earlier, occasions arise where hosts in different VLANs need to communicate with each other over ATM. VLANs have similarities to ELANs on the ATM network. VLANs describe broadcast domains in a LAN environment, whereas ELANs describe broadcast domains in an ATM environment. Whenever hosts in one VLAN need to communicate with hosts in another VLAN, the traffic must be routed between them. If the inter-VLAN routing occurs on Ethernet, Multilayer Switching (MLS) is an appropriate choice to bypass routers. If the routing occurs in the ATM network, MPOA is a candidate to bypass routers. MLS and MPOA both have the same objective: to bypass routers. One does it in the LAN world (VLANs), the other does it in the ATM world (ELANs).

This chapter details the inter-VLAN/inter-ELAN communications performed with MPOA. Without MPOA, the traffic follows the hop-by-hop path described earlier in the chapter.