POINT OF VIEW Common VoIP Architecture

Executive Summary
This white paper describes the architecture of AT&T’s common infrastructure for real-time communications
services over Internet protocol, commonly referred to as VoIP. The infrastructure will be used to
implement all planned and future real-time services that use IP as their main transport technology.
Existing voice services will migrate to this architecture over time and when appropriate.

AT&T is migrating to a single Multiprotocol Label Switching (MPLS)-enabled IP core network in support
of a major initiative, the Concept of One, which will ensure optimum use of resources by maximizing
commonality. All services will eventually be carried over this IP network. The common infrastructure
described in this paper is limited to the essential capabilities and components that will enable AT&T to
develop and deploy real-time services, such as voice, interactive text and video, and various multi-media
collaboration services.

The main goal of the AT&T VoIP architecture is to provide a single, common, and shared infrastructure
that facilitates the development of real-time services with the highest quality and availability, the shortest
possible time-to-market, and the lowest cost of operations and maintenance feasible. To accomplish this,
the architecture is divided into separate independent layers. Each layer has a well-defined role and
provides a set of capabilities for the layer immediately above it by utilizing the set of capabilities provided
to it by the layer immediately below it.

The VoIP infrastructure is built as a virtual network on top of AT&T’s Converged IP/MPLS Core Network
and is called the VoIP Connectivity Layer. The Core Network is surrounded by a Multi-Service
Access/Multi-Service Edge network that supports all popular access technologies including TDM, ATM,
Frame Relay and Ethernet. Thus, the VoIP infrastructure can be reached via any of these access
technologies. Moreover, the architecture provides capabilities to support various access protocols such
as, H.323, MGCP, MEGACO, SIP,TDM/SS7, as well as any VoIP protocol may emerge in the future. This is
achieved by surrounding the VoIP Connectivity Layer with Border Elements (BEs), which mark the trust
boundaries of the VoIP Infrastructure and translate the specifics of various VoIP access protocols into
Session Initiation Protocol (SIP) – the single common internal protocol used by all VoIP infrastructure
components. BEs not only provide protocol conversion, but also enforce various policies including those
needed for call admission control and VoIP-level security.

AT&T Point of View / VoIP – 12/22/03 1

 The Call Control Element (CCE) controls and manages the VoIP infrastructure and provides a single interface to
application servers residing in the Applications Layer. Working with various BEs, the CCE creates and removes call legs
and joins call legs to establish connectivity between end-points. The application servers residing in the Applications
Layer provide the service logic capabilities to implement various types of services. The “plug-in” paradigm is used
to allow easy addition of various services without impacting the VoIP Infrastructure. Vendor-provided application
servers as well as internally developed service logic can both be used in a similar fashion.

Concept of Zero is followed to provide “zero-touch” provisioning and maintenance from a single virtual
operations center.

Introduction to VoIP Layers

Resource Layer
The VoIP archicture is devised in a series of Layers, as illustrated

Operations Support Layer

in Figure 1. The layers follow the principle of information hiding; Applications Layer

that is, each layer has a well-defined role and provides a set of
Connectivity Layer
well-defined capabilities to the layer to which it is sending
BE BE BE BE BE BE
information, by using the capabilities of the layer it received
information from. Core Network Layer

Access

Access

Access

Access

Access

Access
Principles
AT&T’s VoIP Common Architecture Access Layer

• Use a common architecture for all real-time communications services Figure 1:VoIP Layers

• Ensure that high reliability, availability, and performance is comparable

The Access Layer interfaces with customer
to AT&T’s circuit-switched network
equipment and provides customer
• Support both existing and future services connectivity using access-specific protocols.
• Provide for rapid deployment of new services
The IP/MPLS Core Network Layer
• Maximize usage on a single, shared set of resources provides IP connectivity for all elements of
• Accommodate multiple popular access technologies the VoIP infrastructure.
POINT OF VIEW

• Use SIP as the common internal signaling protocol for all signaling The Connectivity Layer provides the VoIP
between components infrastructure needed to process basic calls,
• Leverage the latest, cost-reduced technologies support high performance network
functions, send network primitives, provide
• Allow only Open Standard Protocols for all internal interfaces
media services, interact with Application
• Ensure that components are interchangeable and interoperable Servers for more advanced calls, and
• Design performance and reliability into all components, services support Call Detail Recording. Border
and processes Elements translate the access protocol to SIP.

• Provide for the necessary security to protect AT&T and customer assets The Applications Layer consists of several
• Support “users” in addition to “lines” Application Servers, each providing one or
more services.
• Ensure that capacity closely tracks demand

AT&T Point of View / VoIP – 12/22/03 2

 The Resource Layer provides an environment for the creation of service logic and the management of services
including customer record maintenance and billing plans.

The Operations Support Layer consists of multiple applications, databases and a supporting data communications
network, which are used by internal and external personnel to manage the VoIP network and its elements.

Service Creation
Environment (SCE) Network
Resource

Resource Layer
SLEE XML, CPL, Servlets

App App
Logic Logic Application
Server (AS)

SIP Applications Layer

VXML & HTTP
Common Network Functions
Media Server
Operations Support Layer

Call Admission Network Routing

SIP Control (CAC) Engine (NRE)
E-911 Server
SIP Call Control
Element (CCE) SIP
CALEA Server SIP Service Broker User Profile
SIP (SB) Engine (UPE)
Peer Network SIP
Border Element SIP
SIP SIP Core
Peer IP Network Network Layer H.323 BE
Network
Gateway BE
MEGACO SIP BE
Gateway BE
Connectivity Layer
MEGACO SIP SIP SIP H.323 H.323
Partner IP
SS7 Carrier
Network Managed
Router Managed
Customer Managed Gateway
LS DSL/Cable Managed Router
4E TDM Gateway
LAN

LAN
TA / MTA PBX

PBX PBX PBX

ILEC IP PBX
IP Device

IP Phone IP PBX

IP Device

Access Layer

Figure 2: AT&T VoIP Functional Architecture

POINT OF VIEW

Functional Architecture
In Figure 2, layers are represented in a high-level AT&T VoIP functional architecture. This is a logical functional
architecture and is not intended to represent the physical implementation. The architecture does not specify how
functions will be distributed to devices. Instead, it allows the flexibility to package functions into various servers
using cost-effective methods, as long as open interfaces to those functions are available for use by other functions.

For each access type, a corresponding Border Element (BE) is deployed to manage the access-specific
requirements and interfaces. BEs are the demarcation points for the Connectivity Layer, which is a common
and shared layer on top of the converged IP/MPLS network, or Core Network Layer. The BEs translate access

AT&T Point of View / VoIP – 12/22/03 3

 protocols to SIP, then provide call details to the Call Control Element (CCE). The CCE manages the VoIP
infrastructure and creates, removes and joins call legs. To provide a service, the CCE invokes an Application
Server (AS), using SIP to communicate. By sharing a common architecture, new access technologies can use
all existing and future ASs, and new ASs can support all existing and future access technologies.

This architecture supports all real-time communications scenarios, like prepaid card, click-to-chat, and
teleconferencing. For a simplified basic call flow, the following steps are performed:

1.The caller’s phone connects to the BE using the appropriate access-specific signaling protocol.

2.The BE sends a SIP INVITE to the CCE with Request-URI for the destination’s phone number. The CCE
consults with the Service Broker, which indicates that this call has no features.

3.The CCE sends an INVITE to the destination’s BE, which communicates with the destination’s device using the
appropriate signaling protocol.

4.The call is set up between the two end-points, and the caller and destination talk.

5.The destination hangs up. The BE sends a SIP BYE to the CCE.

6.The CCE sends a BYE to caller’s BE, which disconnects the caller.

Access Layer
The Access Layer provides connectivity between the Customer Premises Equipment (CPE) and the BEs. The
Access Layer must support the endusers’ service requirements such as Quality of Service (QoS), Security, and
Availability. Current technologies will be supported allowing AT&T to continue to offer existing profitable
services at a low cost.

Network MEGACO
The VoIP architecture is designed to Gateway BE BE SIP BE H.323 BE

provide customers with menus of

H.323
3
H.32
services that are independent of
SIP

SIP

Edge SS7
Capability Network
MEGACO

Partner IP Managed
access technologies. In the past, new Carrier Gateway
SIP

4E TDM LS 3
H.32

access technologies were managed

POINT OF VIEW

Gateway / Gateway /
Router Router
AT&T Access
differently. With this architecture, new
access technologies will be deployed Customer / 3rd Customer
Party Access Managed
Gateway
quickly and efficiently without ILEC PBX
DSL /Cable LAN
disturbing the existing VoIP
infrastructure. PBX
TA / MTA
LAN
IP PBX
PBX PBX IP Device

IP Device IP Phone IP PBX

Figure 3: Access Layer

AT&T Point of View / VoIP – 12/22/03 4

 Access Methods
Ubiquitous access to customers is achieved with an Access Layer based on a combination of AT&T and 3rd party
segments. While customers in large enterprise locations can be economically served with dedicated access
(Leased Line, FR, ATM, etc), customers in home offices and small branch offices need to be served by shared
access methods (ILEC Local Service, Cable Modem, DSL, etc). To meet these widely varied requirements, the
Access Layer is composed of a variety of access methods:

• TDM via Edge Switching – Customer access occurs via existing connections to AT&T Edge switches. This
includes shared access from the ILEC and dedicated access to 4E or 5E switches. Bearer and SS7 signaling for
Edge switches interface with the Network Gateway BE. In the future, the Edge switches could support an IP
interface, eliminating the need for TDM to IP conversion in the BE.
• TDM Direct Access – Customer PBX or ILEC switches could interface directly to a Network Gateway BE.
Full range of SS7, ISDN, CAS signaling needs to be supported.
• TDM via IP Direct Access (Managed Router/Gateway) – A direct-access line shared with an AT&T IP Data
Service, (e.g. MIS, MDNS, MRS, EVPN) in conjunction with a VoIP Gateway on the CPE Managed Router, is
used to provide a TDM service demarcation on the customer’s premises. QoS implementation is at
Layer 3 for direct IP Access, and at both Layers 2 and 3 for FR/ATM access.
• IP MIS (Managed Router) – A direct-access line shared with an AT&T IP Data Service, (e.g. MIS, MDNS, MRS,
EVPN) in conjunction with a CPE Managed Router, is used to provide an IP service demarcation for VoIP
service on the customer premises. VoIP protocols supported include SIP for SIP Phones or SIP Proxies and
H.323 for IP PBXs. QoS implementation is at Layer 3 for direct IP Access, with possible additional Layer 2
QoS features for FR/ATM access.
• IP Broadband Agnostic – Customer-purchased Broadband Access, (e.g. Cable, DSL) in conjunction with an
AT&T-provided Terminal Adapter, is used to provide Remote Office or second-line service to individual end
users. End user devices can be SIP Phones, SIP Clients, Black Phones or other IP devices. For Black Phones,
the AT&T Terminal Adapter will convert to VoIP. QoS is achieved by queuing mechanisms implemented in the
Terminal Adapter. Privacy is achieved by IPSec sessions between the Terminal Adapter and BEs.
• IP Peering – AT&T provides or purchases VoIP minutes over an IP Interconnect with another VoIP carrier.
Currently, the VoIP signaling protocol is H.323. Potential relationships with cable operators may introduce
CMSS, a PacketCable variant of SIP. Encryption for Privacy is an option.

Customer Premises Equipment

Customer Premises Equipment (CPE) are end-point VoIP devices at customer locations. Examples include
POINT OF VIEW

gateways, IP PBXs, and IP phones. AT&T selects and certifies CPE to work with AT&T’s VoIP architecture.
CPE interfaces will comply with standard protocols such as SIP, MEGACO, MGCP, and H.323.

CPE Functional Components

Signaling – CPE implement supported access protocols (H.323, MGCP, MEGACO, SIP). CPE are typically the end
point devices for originating or terminating signaling and/or media streams.

Media Control – CPE terminate the media streams going in and out of the Connectivity Layer and provide for
final media format conversion.

AT&T Point of View / VoIP – 12/22/03 5

 Core Network Layer CPE Types

The primary function of the Core H.323 Implements a gateway or terminal function for access.

Network Layer, also known as the

MEGACO/MGCP Implements a MEGACO/MGCP end-point for access.
IP/MPLS Converged Network, is to
provide high performance intersite
SIP Implements a SIP User Agent for access.
routing and transport of IP packets.
In support of the Concept of One, TDM Connects through an NG BE to access VoIP service.
the Core Network Layer is a single
network that provides IP connectivity
for the VoIP infrastructure. The Core Network also has the added value of architecturally isolating the BEs and
the associated Access Networks they connect with, from the CCE, and associated service intelligence. The design
facilitates inter-working among end-points in different access domains. That is, a multi-location customer could
have a single voice-oriented service that works in the same way at all of the customer’s locations regardless of how
network connection is achieved. (e.g.,TDM, IP, ATM, Frame Relay, or remotely via a service partner like a cable company.)

IP Quality of Service Capabilities

In order to address the performance needs of each of the typical traffic streams associated with the VoIP
Architecture (bearer channels, signaling, management traffic, etc.), the Converged IP/MPLS Network supports
tuning of per-hop behaviors within the network. These behaviors are consistent with the performance needs of
four classes of traffic flows, including Real-Time, High Performance Data, Medium Performance Data, and Best-
Effort. This includes setting up priority queues, where appropriate, to support traffic that has real-time
characteristics (i.e., bearer channel traffic). It also includes allocating the remainder of the queue available on those
links to three distinct classes of service (high, medium and best-effort data traffic), with servicing characteristics
existing so that the lower classes cannot overwhelm capacity reserved for the higher classes, when contention
occurs. The equipment signals to which of the four traffic classes to map into by marking IP Precedence settings
within the IP Type of Services (TOS) header (also known as Packet Marking). In the future, the Converged
IP/MPLS Network will support explicit
Admission Control mechanisms.
POINT OF VIEW

Media
CCE
Security Server

By design, the VoIP Architecture is VAR VAR

BR BR
intended to be resilient to known forms
Core Network
of security attacks. This includes both AR BR (trusted edge+core)
simple attempts to steal service, as well
VAR
as brute force distributed denial of SIP Border
VAR

Elements
service (DDOS) attacks intended to AR

dramatically reduce the Quality of Trusted but vulnerable

Service (QoS) provided to customers. portion of edge network
NG Border
Elements Partner Unmanaged
Access IP access
Network

Figure 4: VoIP Security Framework

AT&T Point of View / VoIP – 12/22/03 6

 The definition of a core trust domain where VoIP Architectural Elements participate is fundamental to this design.
Also, the security design assumes that the BE will act as a gateway into that trust domain, from customers who
are natively attached to networks outside of that trust domain (e.g., partner networks, customer premises
networks, etc.).

In the early stages of this architecture, these trust domains may need to be implemented by classifying all of the
traffic on a given interface to be part of a particular type of VPN. In the target implementation, MPLS VPN will be
used, which will both improve the flexibility of VPN management, and allow the components in the other layers of
the VoIP Architecture to mark different traffic streams (via labels) that belong in different VPNs on the same
physical interface.

Connectivity Layer
The VoIP Connectivity Layer provides all network primitives needed for applications to implement services. This
includes establishing simple connectivity between end-points by providing capabilities to create, join, remove (tear
down) and report the status of call legs. The following services are also enabled in this layer: Basic media services,
E-911, CALEA and Call Detail Recording.

The Connectivity Layer provides Connectivity Layer Definitions

a unified, shared environment that Border Element (BE) Translates an access protocol to and from SIP, enforces security,
enforces admission pollicies in coordination with CAC, detects
supports the addition of new DTMF and other types of in-band user requests and redirects
media when instructed by the CCE.
services and access technologies
without changing the basic
CALEA Server Responsible for the interception of communications for law
infrastructure. enforcement and other purposes.

This layer is comprised of a number Call Admission Responsible for the enforcement of network-wide
Control (CAC) admission policies.
of zones and a collection of network
functions. A zone consists of a CCE Call Control Manages and interacts with all BEs in a zone, sets up and tears
managing one or more BEs. Network Element (CCE) down call legs and interacts with ASs to implement their requests.

functions and resources, as described

E-911 Server Routes emergency calls to the appropriate Public Safety
in this section, are shared by CCEs Answering Points, (PSAPs) based on the caller’s location.

across many zones.

POINT OF VIEW

Media Server (MS) Handles and terminates media streams, providing functions such
as announcement, DTMF, text-to-speech (TTS), automatic
Border Element speech recognition (ASR), bridging and transcoding.

The Border Element (BE) is the Networking Routing Provides route information and translates a network address
Engine or logical address to an IP address.
point of demarcation for the
Connectivity Layer. It identifies the
Service Broker (SB) Provides the CCE, upon request, the address of the appropriate
Boundary of Trust for the AT&T VoIP AS to treat the call.

Network and provides an entry

User Profile Contains user data, including registered devices, presence
point into the VoIP infrastructure. Engine (UPE) information and geographical location.

For each customer site, BEs will

Zone Consists of a set of BEs managed by one CCE.
ensure reliability and availability.

AT&T Point of View / VoIP – 12/22/03 7

 As the interface between AT&T’s internal address space and the customer site’s address space, the BE is
responsible for recognizing and authorizing end-points (IP-PBXs, SIP phones, etc.). Functionally, the BE can
be logically decomposed or distributed, if needed. A configuration is supported in which the media-related
processing of the BE takes place at a trusted CPE. The signaling function of the BE always occurs in an
AT&T network location.

BE Functional Components
A BE has four main functions: Signaling, Media Control, Security and Call Admission Control.

Signaling – A BE proxies both the caller and the called end-points, thereby providing a point of signaling control
at the edge. It translates the access protocol (H.323, MGCP, MEGACO, SS7, CAS, ISDN, etc.) to and from SIP.

Media Control – A BE examines all media streams going in and out of the Connectivity Layer for security, media
format conversion, and media transfer. The BE also redirects media streams upon request from the CCE without
impacting the actual caller and called party, and provides the means for the CCE to define, detect and report
DTMF strings during the call.

Some services require the ability for the AS to detect a signal that does not need to be on the media path
(e.g. DTMF, flash hook). For example, a prepaid card application may permit the caller to enter a sequence of
digits (e.g. “**9”). This forces a hang-up of the destination and provides the opportunity for another call to be
placed. To enable efficient utilization of network resources, the BE will allow an AS to register event triggers via
the CCE, and the BE will signal the AS when the event occurs. While there is not currently a standard mechanism
for accomplishing this task, AT&T supports the IETF proposal, using HTTP to send DTMF signals.

Security – A BE provides all necessary security and screening for the customer sites it interacts with. It
authenticates subscribers, customers, and partners, and provides NAT and firewall functions as appropriate.

Call Admission Control – A BE enforces admission policies over the access pipe such as:

• Call gapping – limiting the call

set up rate BE Types

• Call limiting – limiting the Network Gateway Provides all required PSTN interface functions, including SS7
POINT OF VIEW

number and type of calls (NG) signaling and media conversion. The NG BE also operates as a
gateway between circuit-switching and packet-switching networks.
• Bandwidth management – From the CCE, it acts as a SIP User Agent (UA).

ensuring media bandwidth

being sent matches the profile H.323 Implements gatekeeper-routed signaling for H.323 CPE gateways,
negotiated through signaling from the access side. From the CCE, it acts as a SIP UA.

Furthermore, the BE uploads local

MEGACO/MGCP Implements Call Agent (or Media Gateway Controller) signaling
policy information via policy protocol for MEGACO/MGCP devices, from the access side. From the
CCE, it functions as a SIP UA.
or Operations Support, and keeps
track of resources for access SIP Implements SIP Proxy and back-to-back user agent (B2BUA)
functions for SIP devices.
networks.

AT&T Point of View / VoIP – 12/22/03 8

 Call Control Element Call 1
The Call Control Element (CCE) is
UA UA
responsible for providing a call leg view media

of the Connectivity Layer to the ASs. BE - 1 MS

Working with the BEs, the CCE

completely hides the specifics of the UA - 1
Leg 1 Leg 2
Access and Connectivity Layers from
CCE
the ASs; allowing them to deal with
logical end-points (users), call legs, and UA - 2

Service
calls terminating at logical end-points.
An AS can ask the CCE to establish a Call 2
call leg to or from a logical end-point UA
Leg 3
(user), to join call legs to form calls, BE - 1
and to tear down call legs and calls. media

The CCE does not need to interact

UA - 1
with an AS to serve basic calls with
UA CCE
no service features. Leg 1
BE - 2
UA - 2
The CCE functions as a SIP Back-to-
Back User Agent (B2BUA), and is a
signaling end-point for all call legs with Figure 5: Call Leg Relationships

all BEs. Media paths, however, are

established directly between BEs without going through the CCE.

The CCE may receive a “call invocation” (SIP INVITE) from any BE on behalf of a subscriber, or from any server
inside the network. An invocation has information about the caller, the called party, and characteristics of the call
(class of service, media type, bandwidth, compression algorithm, etc.). The CCE may instruct a BE to redirect the
media channels associated with a call to a different destination.

A CCE manages all BEs in its zone. It is aware of the status of each BE it manages, including whether the BE is
POINT OF VIEW

operational or congested. Moreover, the CCE enforces various routing policies, such as deciding which BE to
use to set up a call leg.

The CCE interacts with the Service Broker (SB), the Network Routing Engine, (NRE), the User Profile Engine
(UPE), and the Call Admission Control (CAC). In addition, the CCE also communicates with the resource
servers residing in the Connectivity Layer, such as MS, E-911, and CALEA.

The CCE interacts with the SB to determine the services associated with a call leg (either in-bound or
out-bound). For each call, the CCE provides relevant information and requests the AS to execute the
corresponding service logic.

AT&T Point of View / VoIP – 12/22/03 9

 The CCE interacts with the NRE to translate a network address (such as an E.164 voice network address)
or logical address to an IP address. The CCE uses the physical address of an end-point when considering BE
availability, load, etc. A physical address consists of the IP address of a BE and all other information needed by
the BE to locate the end-point in the address space it manages. Physical address characteristics include:

• A local IP address, when a BE manages an IP address space

• Identification tokens, when a peer network is involved
• The end user’s telephone number for a PSTN

The CCE may also consult with the NRE to learn the terminating BE’s address. If the terminating BE is not in the
zone of the originating CCE, the NRE returns the address of the CCE in charge of the terminating BE. Then, the
originating CCE consults with the terminating CCE, which functions as a SIP redirect proxy, to learn the address
of the terminating BE.

The CCE interacts with the UPE to determine if a user is registered or logged-on, and to determine which
address or number should be used to route the call.

The CCE interacts with the CAC to authorize users and admit calls at the time of call setup, leveraging
network-wide conditions and policies. The CCE formulates the QoS/Bandwidth/SLA policy for network-wide
resources in conjunction with the local policy and customer-specific requirements provided by the BE.

In the future, policy information will be obtained from the lower IP network layer entity, like Bandwidth
Broker/Router, to verify bandwidth is reserved, thus ensuring QoS before acceptance of the call (e.g., ringing will
not be provided unless resources have been reserved).

Common Network
Functions AS
UPE

Common Network Functions used to

provide services across the MS

SIP NRE
Connectivity Layer are the Service SIP
POINT OF VIEW

Broker (SB), the Network Routing SIP

CALEA SIP
Engine (NRE), the Call Admission SIP
SB
Control (CAC), and the User Profile CCE
SIP

Engine (UPE). These functions can be E-911

SIP

used by more than one CCE. SIP Policy AAA CAC

SIP Protocol

SIP
To peer
CCE BE
Policy AAA
Protocol

Figure 6:Architeture of Common Functional Entities in the Connectivity Layer

AT&T Point of View / VoIP – 12/22/03 10

 Service Broker
The Service Broker (SB) maintains subscribers’ service information from a database of installed and activated
services. This database can be shared with other functional entities, such as the NRE and the UPE. The SB acts
as the SIP Redirect server and provides tables that define the services subscribed by each individual user. These
services, along with the associated ASs, are usually provisioned in the SB using Operations Support provisioning
capabilities. In the future, it is expected that any new services and associated ASs will be provisioned with the SB
dynamically, using SIP.

The SB also performs service precedence and identifies to the CCE the primary applicable AS and a list of other
services applicable to the call.

Network Routing Engine

The Network Routing Engine (NRE) provides the route information, upon finding the destination BE. This route
information is required to set up the call leg between the source and destination BEs. SIP is used between the
CCE and the NRE, with the NRE acting as the SIP Redirect server.

When a user registers or subscribes to a service, the database used by the NRE may be provisioned with the
addresses of the end-points, residing in the access network, that are reachable through a particular BE. Also, the
database is populated with a list of BEs through which a user can be reached.

The Location Server is part of the NRE and uses the same database. It collects the telephony route information
(E.164) from the PSTN/TDM-IP gateway when users are reachable through those gateways in the PSTN/TDM
network. The PSTN-IP gateway may be a part of the BE while the TDM-IP gateway may be a part of the CPE.
The following protocols are used:

• TRIP-GW registration protocol is used by the gateways to register and transfer the route information from
the gateway to the Location Server (LS)
• I-TRIP protocol is used between the NRE and the LS, and between the LSs within the AT&T network
• TRIP protocol is used between the LSs of the AT&T Network and AT&T partner or other service provider network
POINT OF VIEW

Call Admission Control

The Call Admission Control (CAC), engaged at the time of call setup, considers network-wide conditions and
policies. The CAC manages capacity, controls congestion, observes firewall restrictions, and interprets SLA, QoS,
Network Address Translation (NAT), and security policies.

The CAC is envisioned as a centralized function controlled by the CCE in collaboration with the BEs. While a
BE implements local and customer-specific policy and provides authentication for entities that access the AT&T
network, CAC functional entities help the CCE determine whether to admit or reject a call request. These
functional entities include the Policy Server and the Authentication, Authorization and Accounting (AAA) server.

AT&T Point of View / VoIP – 12/22/03 11

 The Policy Server is used primarily for SLA and QoS policies. This server may also be used for security, accounting,
billing, firewall, and NATs. The policy protocol used between the Policy Server and other functional entities
(e.g., CCE, AAA Server, or BE) must be based on an open standard protocol that can interwork with SIP, once
appropriate extensions are added to the protocol.

The Authentication,Authorization and Accounting (AAA) Server receives authentication from the BE and uploads
security policy. The AAA protocol used between the AAA server and other functional entities (e.g., CCE or BE) must
be an open standard protocol that can interwork with SIP once appropriate extensions are added to the protocol.

User Profile Engine

The User Profile Engine (UPE) is a functional entity that keeps both static and dynamic user profiles. The UPE
consists of a SIP Registrar and a Presence Server (future implementation), which, along with the NRE, share the
same logical database. Access to the UPE is controlled by the CCE for registration and presence information,
using SIP signaling messages.

The Registration Server receives SIP registration requests under the control of the CCE, via BEs. The user
location information obtained from the registration messages is updated into the user profile database. Users
who subscribe to AT&T services will register with this server under the control of the CCE, via BEs.

The Presence Server (future implementation) stores and conveys the user’s location and status (e.g., offline, busy,
other). Users will register with the Presence Server (PS) under the control of the CCE, via BEs. The PS will
subscribe to the users’ presence information. When a status changes, (e.g., online, offline, busy) notifications will
be sent by users’ presence agents and immediately update the PS. User’s availability should map to presence
information so that the call can be expediently routed to the address where the user is currently available.

Media Servers
Media Servers (MS) typically operate with ASs to handle and terminate media streams, and to provide services
such as announcements, bridges, transcoding, and Interactive Voice Response (IVR) messages. Using SIP to
communicate, the AS sends an invite to the MS, via the CCE, setting up the call and specifying the script the MS
executes or the function it performs. The MS returns the status and results to the AS via HTTP posts. With the
POINT OF VIEW

exception of network-busy announcements provided by NG BEs, components of this architecture will not provide
their own MS functions.

Special Network Servers

The following are special network servers used in the VoIP network.

CALEA Servers provide the ability to identify and collect content of voice telephone calls traversing the
VoIP network, as mandated by the Communications Assistance for Law Enforcement Act (CALEA). When a
CCE detects that a call needs to be monitored for legal reasons, it sets up the call via the CALEA server.

E-911 Servers route calls to the appropriate Public Safety Answering Points (PSAPs) based on the caller’s
location. When the CCE receives an emergency 911 call, it sends the call to the E-911 server and ensures
that call waiting is disabled for the duration of the call.

AT&T Point of View / VoIP – 12/22/03 12

 Call Detail Recording
Call Detail Recording is required to provide data for billing and operations.The target is to have all BEs, CCEs, and
ASs generate Call Detail Records (CDRs) for Operations Support, and to have the CCE generate an integrated
CDR for billing purposes.The BEs and ASs should send the required information for billing to the CCE, via call
signaling. There are several key requirements that will enable state-of-the-art network management and billing:

• Near-real time delivery to traffic monitoring, fraud detection, and billing systems
• Vendor-independent formats based on industry or standards-based formats
• Inclusion of quality of call and routing information
• Capturing PSTN interconnection and termination release codes
• Capturing a Unique Call ID from the originating BE
• Clearly-defined audit methods
• Record storage and redelivery in the case of lost connectivity

Applications Layer
Application Servers (ASs) implement features that leverage the VoIP Architecture, using a common interface
to the VoIP Connectivity Layer. The “plug-in” paradigm facilitates fast and easy addition of new features, and
SIP creates a unified interface to the Connectivity Layer.

“Plug-In” Features
The CCE, as directed by the SB, uses SIP to invoke an AS.The AS may also initiate communication sessions via
the CCE (third party call control). While each AS is responsible for implementing its own service logic, all ASs
implement a common interface to the CCE, by which the CCE may invoke an AS, and by which an AS may ask
the CCE to set up, tear down, modify, or join call legs. This interface is implemented using SIP, thus providing
common SIP signaling throughout the VoIP Architecture. ASs are free to implement only the subset of this
interface that they require in order to perform their functions. While this interface may evolve over time, it will
maintain backwards compatibility so existing ASs will continue to function, as the interface evolves.
POINT OF VIEW

While SIP provides the flexibility to allow ASs to set up call legs without leveraging the CCE, in the AT&T VoIP
Architecture, ASs must use the CCE to set up, tear down, modify, and join call legs. Thus, from the AS’s point of
view, each AS has a CCE acting as its SIP Proxy for all inbound and outbound SIP messages.

While there is a well-defined SIP-based interface between the CCE and all ASs, each AS may invoke other Network
Resources, using the appropriate protocol for that resource. Thus, an AS implements an abstract that enables new
feature “plug-ins” without having to introduce new capabilities or protocols to the core VoIP Architecture.

For example, a voice mail application may leverage an IMAP message store and an LDAP directory, both of
which are examples of Network Resources. Network Resources need not be operated or even hosted by the
VoIP service provider, as in the case of a “Find Me” application leveraging a user’s calendar that is available over
the Internet as an XML Web Service.

AT&T Point of View / VoIP – 12/22/03 13

 As shown in the diagram, the AS may
Customer-
also invoke the MS to play prompts, Network Network Network
Owned
Resource Resource Resource
Resource
collect user input, etc., by asking the
CCE to create a call leg to the MS via
SIP. The CCE will use SIP to create the
leg to the MS. The MS will then Application
Servers
request VoiceXML scripts and audio
files, and post back user input via
HTTP, as defined by the VoiceXML 2.0 HTTP HTTP
SIP
standard. For efficient use of network
resources, an AS may ask the CCE to
SIP SIP
Media Server
instruct a BE to notify the AS when (VXML based) CCE BE

it detects mid-call DTMF.

Figure 7:Application Server Interface

Service Creation
While many ASs implement specific service logic as defined by the vendor, certain ASs will instead serve as
software platforms for implementing new services. These ASs have two elements: Service Creation Environment
(SCE) and Service Logic Execution Environment (SLEE).

The Service Creation Environment (SCE) allows a development team to develop and test new service logic
scripts. These scripts integrate various building blocks into a service, where the building blocks may be
precompiled, implemented by other SIP-based ASs, or implemented as Network Resources.

The Service Logic Execution OSS

Environment (SLEE) provides the Web-based service Service Creation

subscription and Environment
run-time environment for executing customization
Network
Resource
service logic scripts. While the SLEE XML, CPL,
Servlets
appears to the rest of the VoIP
Architecture to be just another AS, App
Logic
App
Logic
SIP, SOAP,
LDAP,
SMTP, TCAP
it plays an important role in the User &
SLEE
POINT OF VIEW

Service Profiles
ability to deploy new services quickly,
HTTP HTTP
leveraging the capabilities of the SCE,
SIP App
Server
as well as web-based management
MS BE
of subscribers and web-based service
customization.
SIP
CCE

Figure 8: Service creation and Execution Environment

AT&T Point of View / VoIP – 12/22/03 14

 Resource Layer
The Resource Layer contains all systems that various ASs may use to provide their services. Most of these resources
are owned and operated by AT&T, but a few may be owned and operated by AT&T customers. Protocols are not
defined as a part of this architecture.

Operations Support Layer

The Operations Support Layer consists of multiple applications, databases, and a supporting data communications
network used by internal and external personnel to manage the VoIP network and its elements. The target Operations
architecture supporting the VoIP Network relies on the following design principles:

• Concept of One – Evolve to one Virtual Operations center from which all network elements can be monitored
and managed. Accordingly, consolidations toward one organization, one network, one operations support
architecture and one process result in reduced costs and increased efficiencies.
• Zero Touch – Minimize manual
operator touch of the network Web based tools
elements for all network
External Users Internal Users
management functions. Requires
an accurate network inventory to
support process automation. Network & Service Data
Customer-specific Data

• e-Enablement – Provide access to

network operations support
systems via administrated
permission to web interfaces, in
order to facilitate the Concept of
One and Zero-touch principles.
The native Element Management
Configuration
Management

Management

Management

Management

Management
Performance
Provisioning

System (EMS) provides access for

Capacity

Fault
the work centers into the VoIP
infrastructure components
via a single step cut-through.
• Open Standards – Allow multiple Open Interfaces
vendors to interoperate, and
EMS
AT&T to adopt off-the-shelf
solutions and processes to meet
POINT OF VIEW

business objectives. In general,

MS
suppliers are expected to meet
E-911
open interface standards with their CALEA
VoIP network elements. AS
BE
IP/MPLS
The following Operation Support Network Functions CCE
systems and processes are described SB CAC

in this section: Fault Management,

UPE NRE
Configuration Management, Network
and Service Provisioning, Capacity Converged VoIP Network
Management, and Performance and
Incident Management. Figure 9:VoIP Operations Architecture

AT&T Point of View / VoIP – 12/22/03 15

 Fault Management
All VoIP network elements are External Internal
Users Users
supported through a single Fault EMS Converged
VoIP
Management process. This includes Network
both end-customers and VoIP HTML
Fault Mgmt
infrastructure components (BEs, Data Network
Management
Servers, Engines, Databases, CCEs, Protocol (e.g. SNMP)

CACs, etc.). Trouble

Tickets

Faults from VoIP infrastructure

Traps,
components are autonomously Cause Codes, etc.

reported to the Fault Detection and Alarms

Correlation
Rules Management Systems through
an Open Standard Interface (e.g.,
SNMP). Fault indications provided by
Event
the VoIP infrastructure components Collector

will support unique trap Console

Ports, CLIs
identifications and cause codes. etc.

Terminal
The VoIP infrastructure components Server

support end-to-end testing across

the converged networks from a Figure 10: Fault Management Architecture

common test platform.

Configuration Management
Configuration Management supports the initial loading of element parameters, changes in element configuration,
and retention of data for recovery for all network elements (BE, Server, Engine, Database, CCE, CAC) that are
physically connected to the VoIP Converged Network.

Configuration Management refers to the core operations process of loading the following types of data:
POINT OF VIEW

• Data elements needed to uniquely identify the component and to establish connectivity with other IP
network elements and operations systems
• Data elements that are static and populated at the time of installation
• Ongoing loading of data triggered from facility, trunk (group), link, or routing orders

Configuration Management must support the following:

• Retention of golden copy configurations and a configuration validation process (where configuration data
is stored in the network element)
• An interface with the Network Inventory Database to maintain growth changes in the overall VoIP Network
topology, in concordance with the Capacity Management process.

AT&T Point of View / VoIP – 12/22/03 16

 • The test & turn-up process at equipment deployment
• Showing the configuration and provisioning architecture for the VoIP systems. Appropriate open interfaces
must be supported.

Network Provisioning
Network Provisioning is the process External Internal
Users Users
that assigns previously installed
and configured VoIP network
element capacity (e.g., NG BE) Network
Inventory
to support a level of service. EMS Converged
VoIP
Network and Service Provisioning Network
Golden Configuration
are interdependent, as Service Configuration Mgmt

Provisioning provides the element’s

Network
Management
services. Protocol
Service-
specific data
Network Provisioning of the VoIP
Order
infrastructure components requires Mgmt
Provisioning
standards-based machine-to-machine Mgmt

interfaces to both a transaction Customer- FTP, XML

specific data
initiator, (e.g., legacy OSS) and the
actual infrastructure element being Service
Mgmt
provisioned. Infrastructure elements
must be able to support provisioning SQL Replication

requests from multiple clients for

changes to the state of a customer
Figure 11: Configuration and Provisioning Architecture
(e.g. in-service, out-of-service).
In general,VoIP infrastructure
components should support direct or indirect (e.g., via EMS) access for automated provisioning via SNMP, FTP
and/or database replication tools as necessary. If an EMS is used, it must be able to acknowledge receipt of
POINT OF VIEW

transactions and report successful completion of transactions, or reason for failure, to upstream provisioning systems.

Service Provisioning
VoIP Service Provisioning requires a flow-through mechanism to provision customer-specific orders within the
Converged VoIP network, as well as orders to provision facility, trunking, and routing associated with access to the
VoIP converged Network.

In adherence to Concept of Zero, the VoIP infrastructure elements will support flow-through of customer service
requests by allowing the following:

• Real-time system interfaces to SB, allowing update and query of the services or applications (8YY, SDN, etc.)
that a customer has purchased

AT&T Point of View / VoIP – 12/22/03 17

 • Real-time system interface to service ASs, allowing update capabilities and query of the features for a
given application
• Real-time system interface to the VoIP Service Provisioning Server for network provisioning
• Real-time update of the UPE
• Real-time system interface to billing systems

Capacity Management
Capacity Management is the primary External Internal
Users Users
tool used to maintain adequate
EMS Converged
in-service inventories of all VoIP VoIP
Network
Network Elements, (BEs, Servers,
Network Controls via
Engines, Databases, CCEs, CACs, Fault Network Management Protocol
Mgmt
Data
etc.) in order to meet customer
demands of the network. AT&T
PM Data
uses extensive performance
PM and
measurement data on the network IM Tool
From Fault
elements, including average and peak Mgmt

buffer utilization, as well as packet Call Irregularities,

QoS Data, NE
Utilization, Traffic
loss. Using these inputs, along with Measurements &
Alerts, etc.
service forecasts, the system will be
used to forecast the network’s
physical growth needs.
CM Data Capacity
Mgmt
Performance Management
and Incident Management
Operational support for Incident
Figure 12: Performance and Capacity Management Architecture
and Performance management is
crucial to ensuring AT&T’s VoIP network
POINT OF VIEW

maintains the highest performance levels. Direct and indirect control of people, process and network is
coordinated in these functions, with the goal of increasing service quality. The functions of Performance and
Incident Management can be divided into four functions: Performance Surveillance, Incident Management,
Traffic Management and Performance Reporting.

Performance Surveillance functions require near real-time availability of network behavior measurements
(e.g., faults, call processing irregularities, traffic measurements, equipment alarms, quality of service information
(delay, loss, jitter) and configuration changes. Measurements are analyzed against configurable and perhaps
dynamic thresholds. Resultant performance surveillance information is used to detect and locate problems
that degrade service. Measurements also provide feedback on the effectiveness of restoration and traffic
management activities.

AT&T Point of View / VoIP – 12/22/03 18

 Incident Management requires information about the customer impact resulting from abnormal network behavior
during and following service-affecting events. This requires the ability to relate service degradation with customer
impact in terms of service, to identify which customers were impacted, and to determine the length of the event.

Traffic Management is the practice of real-time traffic control to protect network resources from overload, or
to redirect traffic to make better use of under-utilized network resources. Either goal serves to maximize the
effectiveness of the network. Traffic management is most often associated with mass calling events and national
crises. This discipline allows for prioritization of restorative activities to reflect business imperatives.

Performance Reporting functions are supported by data extracted from stored aggregate performance
measurements. This data needs to be reported by time series (trending), by event, by customer, by service, by
network device or by any ad-hoc extract, for the purposes of quality assurance, process improvement, equipment
maintenance, network configuration, network planning, and business evaluation or customer inquiry.

Performance and Incident management are interdisciplinary, multi-service and centralized functions. OS solutions
from multiple vendors for multiple services need to interoperate. OS support requires centralization and high
levels of integration. Open APIs and standardized system interfaces are needed for network elements, EMSs
and Network Management Systems (NMSs).

Performance and Reliability

The network platforms that support the VoIP Operations Support Layer, which include Operation Support
Systems and their Data Communication Networks, must deliver appropriate contributions to performance
and reliability to ensure that the objectives of all services and operations functions can be met.

AT&T has employed various layers of protection to support survivability by preventing and, when required,
restoring the network/service with minimal customer impact. These layers of protection include outage
prevention, equipment protection switching, facility restoration, routing resiliency, network management and
disaster recovery.

Conclusion
This document has illustrated the high-level concepts of AT&T’s Common VoIP Architecture. The architectural
POINT OF VIEW

platform will enable AT&T to develop and deploy existing and new real-time communication services that use
IP as their transport mechanism. By sharing a single, common VoIP infrastructure with interoperable network
elements that support open standard interfaces and protocols, AT&T will be able to develop new services and
applications, and interconnect with various access networks.

AT&T is committed to creating a high-capacity VoIP network that ensures reliability, availability and performance.
In adherence to the Concept of One, a single infrastructure will be developed to support existing and new
services. And, in adherence to the Concept of Zero, intelligence will be built into AT&T’s core network, providing
little or no need for human intervention. The AT&T Common VoIP Architecture creates a foundation for
the complex and intricate process of migrating voice and other real-time communications services from a
circuit-switched to an IP packet-switched network.

AT&T Point of View / VoIP – 12/22/03 19

 Acronyms MIS – Managed Internet Services
MPLS – Multiprotocol Label Switching
AAA – Authentication, Authorization and Accounting
MRS – Managed Router Service
API – Application Program Interface
MS – Media Server
AR – Access Router
MTA – Media Terminal Adapter
AS – Application Server
NAT – Network Address Translation
ASN – AT&T Switch Network
NG – Network Gateway
ATM – Asynchronous Transfer Mode
NMS – Network Management System
B2BUA – Back-to-Back User Agent
NRE – Network Routing Engine
BE – Border Element
OS – Operating System
BR – Backbone Router
OSS – Operational Support System
CAC – Call Admission Control
PBX – Private Branch Exchange
CCE – Call Control Element
PS – Presence Server
CDR – Call Detail Record
PSTN – Public Switched Telephone Network
CLI – Command Line Interface
QoS – Quality of Service
CMS – Call Management Server
SB – Service Broker
CMSS – CMS Signaling
SCE – Service Creation Environment
CPE – Customer Premises Equipment
SIP – Session Initiation Protocol
DDOS – Distributed Denial of Service
SLA – Service Level Agreement
DSL – Digital Subscriber Line
SLEE – Service Logic Execution Environment
DTMF – Dual Tone Multi Frequency
SNMP – Simple Network Management Protocol
EMS – Element Management System
TA – Terminal Adapter
ES – Edge Switch
TDM – Time Division Multiplexing
EVPN – Enhanced Virtual Private Network
ToS – Type of Service
FR – Frame Relay
TRIP – Telephony Routing over IP
FTP – File Transfer Protocol
UA – User Agent
GW – Gateway
UPE – User Profile Engine
HFC – Hybrid Fiber Coaxial
VoIP – Voice over Internet Protocol
IETF – Internet Engineering Task Force
VPN – Virtual Private Network.
ILEC – Incumbent Local Exchange Carrier
POINT OF VIEW

IMAP – Internet Message Access Protocol

IP – Internet Protocol
OSPF – Open Short Path First
ISP – Internet Service Provider
I-TRIP – Intra-domain TRIP
IVR – Interactive Voice Response Message
LDAP – Lightweight Directory Access Protocol
LS – Location Server
MDNS – Managed Data Network Service
MEGACO – Media Gateway Control
MGCP – Media Gateway Control Protocol

AT&T Point of View / VoIP – 12/22/03 20

 DISCLAIMER:
This document presents information about the current plans and considerations for AT&T’s VoIP Infrastructure. This
information is subject to change without notice to you. Neither the furnishing of this document to you nor any
information contained in this document should be interpreted by you as a legal representation, agreement, or commitment
by AT&T concerning (1) any information or subject contained in or referenced by this document, or (2) the furnishing of
any products or services by AT&T to you, or (3) the purchase of any products or services by AT&T from you, or (4) any
other topic or subject. AT&T may own intellectual property which relates to the information contained in this document.
Notwithstanding anything in this document to the contrary, no rights or licenses in or to this intellectual property are
granted, either expressly or impliedly, by this document. Rights to AT&T intellectual property may be obtained only by
express written agreement with AT&T, signed by AT&T’s Chief Technology Officer (CTO) or the CTO’s authorized designate.

Copyright 2002 AT&T. All rights reserved.

The information contained in this document represents current considerations of the AT&T Common VoIP Infrastructure
as of the date of publication. The contents of this document should not be interpreted to be a commitment or legal
representation on the part of AT&T to you regarding AT&T’s VoIP Infrastructure, and are subject to change without notice.
AT&T MAKES NO WARRANTIES, WHETHER EXPRESS OR IMPLIED, IN THIS DOCUMENT. BY PUBLISHING THIS
PAPER AT&T IS NOT GRANTING TO YOU, EITHER EXPRESSLY OR IMPLIEDLY, ANY RIGHTS UNDER ANY OF ITS
INTELLECTUAL PROPERTY.

The AT&T logo is a registered trademark of AT&T Corp. in the United States and other countries.

For more information, contact your AT&T

Representative, or visit www.att.com/business.

POINT OF VIEW

AT&T Point of View / VoIP – 12/22/03. © 2003 AT&T. All Rights Reserved. Printed in U.S.A. AB-0139 21