A typical end-to-end ADSL services
architecture is illustrated in Figure .
It consists of customer premises equipment (CPE) and supporting
equipment at the ADSL point of presence (POP). Network access
providers (NAPs) manage Layer 2 network cores, while NSPs manage
Layer 3 network cores. These roles are divided or shared among
incumbent local exchange carrier (ILEC), competitive local exchange
carrier (CLEC), and Tier 1 and Tier 2 Internet service provider
(ISP) businesses. It is expected that over time, market forces will
redefine current relationships between ADSL providers: some NAPs may
add Layer 3 capabilities or extend service across the core.
CPE represents any combination of
end-user PCs or workstations, remote ADSL terminating units (ATU-Rs),
and routers. For instance, a residential user may have a single PC
with an integrated ADSL modem on a peripheral component interface
card, or perhaps a PC with an Ethernet or universal serial bus (USB)
interface to a standalone ADSL modem (the ATU-R). In contrast,
business users will more often connect many end-user PCs to a router
with an integrated ADSL modem or a router plus ATU-R pair.
At the ADSL POP, the NAP deploys one
or more DSL access multiplexers (DSLAMs) servicing the copper loops
between the POP and CPE. In a process called subtending, DSLAMs can
be chained together to enhance ATM pipe utilization. DSLAMs connect
locally or via an inter-central office (CO) link to a local access
concentrator (LAC) that provides ATM "grooming," PPP
tunneling, and Layer 3 termination to local or cached content.
A service selection gateway (SSG) may be colocated with the LAC, so
customers can dynamically select destinations. From the LAC/SSG,
services extend over the ATM core to the NSP or IP network core.
As you can see from Figure ,
three different architectures are applicable to wholesale ADSL
services:
- ATM point to point--cross-connects
subscribers to their ISP or enterprise destination with
permanent virtual circuits (PVCs) from the CPE to the endpoint
- Aggregation--aggregates
multiple subscriber virtual circuits (VCs) into trunk PVCs to
reduce the number of VC connections across the network core;
instead of one VC per subscriber, this uses one VC for many
subscribers to the same destination
- SVC and MPLS--uses switched
virtual circuits (SVCs) to autoprovision connections from the
CPE through the DSLAM to an edge label switch router (edge LSR),
where it enters the Multiprotocol Label Switching (MPLS)-enabled
network core.
Figure
outlines the end-to-end protocol stack used with xDSL.
|