Information Technology Reference
In-Depth Information
The result of this specification is that each network element must be able to forward packets
to adjacent network elements via the DCC. The NEs must act as ISs instead of ESs, which
requires them to be part of the Layer 3 topology. Therefore, they must participate in routing.
DCN Architecture
The common Layer 3 architecture in the DCN is for each ring to be its own Level 1 (L1)
area. The NE that connects to the management network is called the gateway network
element (GNE). Smaller rings may have only a single GNE; however, it is common for
larger rings to have two GNEs to provide redundancy.
The management router is a Level 1/Level 2 (L1/L2) router that is responsible for aggre-
gating connections to the GNEs and providing connectivity to the L2 backbone. The man-
agement router can connect to multiple areas using the IS-IS multiarea (IS-IS MA) feature.
If the management router does not support the IS-IS MA feature, a separate L1/L2 router
is required for each ring. The management router connects to the GNE via an L1 adjacency
and connects to L2-only aggregation routers via an L2 adjacency.
Typically, low-end routers are used in the DCN because of their small form factor and the
scarcity of rack space in the CO. This limits the size of the L2 network to approximately
200 routers, or about 100 to 150 COs. It is not uncommon for a single network to consist
of several hundred COs.
This scalability constraint results in the creation of multiple routing domains, which must
all be interconnected. Typically, a core network is built and the L2-only aggregation routers
are connected to the core network using static routing or Interior Gateway Routing Protocol
for CLNS (ISO-IGRP). Static routing and ISO-IGRP do not provide the scalability required
for today's largest DCNs and certainly will not scale to support future networks that might
result from organic network growth or mergers between current DCNs. It was to solve this
scalability issue that support was added to BGP for CLNS routing information. The use of
BGP for CLNS routing scales better than static routes from an administrative perspective
and better than ISO-IGRP from a prefix advertisement perspective.
BGP-Based DCN Network Design
The implementation of CLNS support in BGP relies on TCP, not Transport Protocol 4 (TP4),
to provide the transport layer connection. Peering sessions are built between IP addresses,
not Network Service Access Points (NSAPs). This adds complexity regarding the BGP next
hop for NSAPs, which is covered in the section “BGP Next-Hop for CLNS Prefixes.” This
also necessitates the introduction of IP into the network to let the TCP sessions for BGP
form.
Search WWH ::




Custom Search