TABLE OF CONTENTS
Basic Video Applications and Architecture 2
IGMP and STB Signaling 3
Video Traffic Characteristics 3
Requirements for Success 4
Business 4
Technical 4
Second Mile Aggregation 4
ATM Characteristics 5
IP/Ethernet Characteristics 5
Quality of Service 6
ATM QoS 6
IP QoS 6
The Access Network 6
IGMP and ATM DSLAMs 7
IGMP and IP/Ethernet DSLAMs 7
Multicast Performance 7
ATM Multicast Replication 8
Multicast In IP/Ethernet DSLAMs 8
Conclusion 9
Allied Telesyn’s Solution 9
About Allied Telesyn, Inc. 9
INTRODUCTION
Great changes have hit the telecommunications industry. Telephone service providers find themselves
having to shift business strategies to continue their growth. Faced with fierce competition and regulatory
uncertainty, service providers are looking at video delivery to increase revenues and deflect competition.
However, profitable video delivery requires carriers to think carefully about how best to invest in their
network infrastructure.
ADSL bandwidth advances enable the delivery of a “video pipe” over existing copper infrastructure. New
video compression technologies significantly reduce the bandwidth requirements for digital-quality video
transmission. The combination of these technologies makes video viable for most service providers.
However, carriers must balance decisions about infrastructure technology to address customer needs as well
as return on investment.
Because the value of a network is proportional to the cost of delivering services, today’s network decisions
will determine tomorrow’s revenues. This paper compares two access and core infrastructures for delivery
of video over existing copper: Asynchronous Transfer Mode (ATM) and Internet Protocol (IP).
Multicast Data
Video Signals
ADSL
ATM or IP Core
Video Encoder Set-top Box
DSLAM
Unicast Data
TV TV
The video head-end sources signals via satellite or off air channels. The signals are injected into video
encoders, which convert the analog or digital signal into a 3Mbps data stream. The two most common video
encapsulation formats output by an encoder are MPEG-2 over ATM or MPEG-2 over IP.
The core network delivers video content from the head-end to the appropriate end points, where ATM or
IP Layer 3 switches aggregate traffic from DSLAMs. The core network must transport two specific types of
traffic: multicast and unicast, corresponding to broadcast and video-on-demand (VoD) services. In essence,
the core network must assist the multicast application with distributing and replicating the multicast video
streams by passing such traffic through the core from node to node.
Each node along the path makes a decision about which multicast traffic must be replicated and forwarded
to network links or access interfaces, depending on the topology involved.
Replication is the first point at which ATM and IP infrastructure functionality diverge in video applications.
ATM is a non-broadcast multiple access technology. It cannot benefit from the facilities offered by a broadcast
medium such as IP/Ethernet. In IP, broadcast is a simple function. In ATM, broadcasting requires the creation
of cascading, point-to-multipoint Permanent Virtual Circuits (PVCs) at each ATM switch point. As we will
see, ATM and IP will both scale to deliver multicast, but ATM is not as efficient or cost-effective.
The DSLAM is the last element in the access network before the subscriber’s loop. It is responsible for serving
the video stream to the Set-Top Box (STB) in the subscriber’s home. The STB decodes audio and video
streams and delivers the decoded content to the customer’s audio and video appliances. In addition, the STB
may incorporate a web browser to facilitate Internet navigation, Video on Demand, or Pay-per-View events.
PAGE 03 of 09
Switch queries
for video requests
Layer 2/3 switch
DSLAM
TV
Let’s evaluate the relative capabilities of IP and ATM in the aggregation and access Layers
of a distribution network.
ATM Characteristics
In the ATM architecture, increasing capacity decreases operational efficiency due to the complexity of
managing ATM adaptation layers and PVCs. ATM network bandwidth is burdened by the ATM cell tax
which consumes at least 10% of capacity in overhead. ATM DSLAMs were not designed to handle the
interaction of IGMP protocol signals, which are a requirement for video channel-changing from STBs.
ATM Core
ATM DSLAMs
ADSL
Set-top Boxes
TV TV TV
IP/Ethernet Characteristics
By using IP and Ethernet for aggregation, capital and aggregation costs decrease as subscriber port densities
increase. Operational efficiency increases due to IP’s simple provisioning and because there is no need to
manage PVCs. Bandwidth availability increases because there is no ATM cell tax. Most important, IP is
multicast-oriented because IGMP is a designed-in feature.
IPIP Core
Core
IP DSLAMs
ADSL
Set-top Boxes
TV TV TV
QUALITY OF SERVICE
Quality of service is a crucial requirement in a video delivery system, and ATM and IP deliver sharply
different benefits when providing video-class QoS.
ATM QoS
ATM is noted for its strong QoS. To support video multicast, ATM uses a Constant Bit Rate (CBR) service
to ensure low transfer delay. CBR allocates bandwidth to an application even if no traffic is present. For
intermittent, bursty traffic, ATM typically uses the more bandwidth-efficient Unspecified Bit Rate (UBR)
service.
MULTICAST PERFORMANCE
Another unique IGMP-related function of the DSLAM affects bandwidth availability on the ADSL link.
The DSLAM must have the capability to turn the multicast streams on and off very rapidly, in less than 100
milliseconds, to prevent the ADSL line from being flooded with multicast packets when there is a channel
change. As a result, the DSLAM must support multicast replication of packets or cells in hardware and at
wire speed. Consider the following two scenarios, in which we examine ATM DSLAMs that do not support
internal multicast replication.
ATM Fabric
Figure 8: Multicast replication in an ATM DSLAM using one PVC per STB video stream
PAGE 08 of 09
The second scenario shows the latest generation of ATM DSLAMs, which commonly uses “Abacus-style”
ATM switching fabric. The switching fabric places the first multicast stream on an ATM bus that distrib-
utes all cells originating from the ingress interface card to all line cards. The second multicast bus is a
spatial multicast (cell marker) that replicates cells to the ports on line card. (See Figure 9.)
Second Stage
ATM Bus
Figure 9: ATM DSLAM with one ATM PVC per video stream using first and second-stage busses
The first stage ATM bus does not replicate cells on the ATM bus. Instead it provides every cell with a special
marker, identifying it as a multicast cell with a multicast session ID that addresses a special multicast queue
belonging to a particular STB session. The ATM bus device on the line card recognizes this cell as a
multicast cell and only forwards it, if a multicast session for this particular connection is established. Many
ATM DSLAMs that use this multicast technique may require a second network interface for unicast PVC
traffic, such as Internet service.
Multicast In IP/Ethernet DSLAMs
In contrast, let’s examine the internal working of IP/Ethernet DSLAM. In Figure 10, the heavy horizontal
line represents one IP multicast stream from CNN, arriving on a Gigabit Ethernet interface. Once the mul-
ticast stream enters the DSLAM, the switching logic determines the forwarding path to a particular port.
This is accomplished by looking at the MAC forwarding database table that was created when the STB
requested the CNN video stream.
In the example, we see three subscribers watching CNN. The important advantage is the IP switching
fabric’s ability to replicate and direct one incoming multicast stream to many subscriber DSL ports. The
DSLAM must support multicast replication of packets in hardware and at wire speed. This last mile pack-
et replication by the IP DSLAM is key to a scalable, cost-effective access solution.
PAGE 09 of 09
CONCLUSION
The objective of a Video over ADSL architecture is to cost-effectively deliver multicast and unicast data
across core, aggregation, and access networks. ATM is a connection-oriented technology that requires
point-to-point virtual circuits at each DSLAM for any service. ATM’s point-to-point nature, while useful
in delivering predictable quality of service, limits a service provider’s ability to cost-effectively scale the
delivery of video.
In contrast, IP and broadcast video are similar in their broadcast nature. In both IP and video, all nodes on
the network receive all signals, and it is up to the node to decide what traffic belongs to that node. It is this
similarity of architecture that makes the transition from ATM to IP on the DSLAM a watershed event.