DELIVERING ECONOMICAL IP-VIDEO OVER DOCSIS BY BYPASSING THE M- CMTS WITH DIBA Michael Patrick Connected Home Solutions

Abstract CMTS core, DIBA is an architecture for delivery of high-bandwidth entertainment DOCSIS IPTV Bypass Architecture video over DOCSIS to the home for a price (DIBA) DIBA refers to any of a number of per program that matches the cost for techniques whereby downstream IPTV traffic conventional video over MPEG2 delivery to is directly tunneled from an IPTV source to a set-top-boxes. downstream Edge QAM, “bypassing” a DIBA is an innovative bypass DOCSIS M-CMTS core. It will allow architecture in which the conventional operators to deliver economic IP video traffic Modular-CMTS architecture is replaced by a over DOCSIS infrastructure. hybrid architecture consisting of an integrated CMTS and DOCSIS External Physical INTRODUCTION Interface (DEPI) or MPEG Edge QAMs. This paper/presentation will demonstrate the MSOs are considering IP-video and IP effectiveness of DIBA, discuss how operators Television (IPTV) to supplement their current can evolve their infrastructure to implement digital video delivery. IP-based video enables DIBA, discuss alternative bypass new video sources (the Internet) and new encapsulation protocols to tunnel traffic to video destinations (subscriber IPTV playback either DEPI or MPEG Edge QAMs, and devices). outline the economic advantages of DIBA Of course, such a transition to IPTV migration. It will explain how DIBA removes requires significant additional downstream the boundaries of delivering video using a M- DOCSIS® bandwidth. In a reasonable 5-7 CMTS, and how operators can accelerate year “endgame” scenario, VOD is expected to service velocity by efficiently delivering require 26 times the bandwidth of High Speed IPTV services that bypass M-CMTS Data (HSD). What’s more, attempting such an platforms. increase in downstream bandwidth with a conventional Modular-CMTS (M-CMTS) will THE CHALLENGE OF DELIVERING IPTV be expensive. As an alternative, we propose that IP- IP-video and IP Television (IPTV) create video content travel directly to an Edge QAM, opportunities to supplement their current bypassing the M-CMTS core altogether and digital video delivery. IP-based video enables saving local switch and CMTS core capacity. new video sources (the Internet) and new This technique is called the DOCSIS IPTV video destinations (subscriber IPTV playback Bypass Architecture (DIBA). DIBA refers to devices). It should be considered as a delivery any of a number of techniques whereby option of Video on Demand (VOD). downstream IPTV traffic is directly tunneled The extent to which VOD is delivered as from an IPTV source to a downstream Edge IPTV governs the extent to which additional QAM, “bypassing” a DOCSIS M-CMTS core. downstream DOCSIS bandwidth must be By bypassing the high-cost components of the provided for IPTV. Consider a hypothetical 5- 7 year “endgame” scenario where each video subscriber can download “What I Want When I Want” (WIWWIW). Assume a fiber node of AN INTRODUCTION TO DIBA 750 homes passed has 75% video subscriber penetration, and during the busiest video But there is a simple alternative for IPTV viewing hour 50% of the video subscribers support with a M-CMTS: IPTV should be require an average of 10 Mbps VOD content. tunneled directly to the EQAM, bypassing the The total VOD content required to be M-CMTS core altogether. delivered to that single fiber node during the DIBA refers to any of a number of busy hour is thus 750 * .75 * .50 * 10 Mbps = techniques whereby downstream IPTV traffic 2.8 Gbps, or 73 carriers of 6 MHz QAM 256. is directly tunneled from an IPTV source to a What fraction of this “endgame VOD” downstream EQAM, “bypassing” a DOCSIS bandwidth is IPTV over DOCSIS and what M-CMTS core. By bypassing the high-cost fraction is traditional Digital Video to Digital components of the CMTS core, DIBA is an Set-Top Boxes (DSTBs) is anybody’s guess. architecture for the delivery of high- VOD bandwidth swamps High-Speed bandwidth entertainment video over DOCSIS Data (HSD) bandwidth. Last year, most MSO to the home for a price-per-program that deployments offered only 20 Mbps of HSD matches the cost for conventional video over per 750-household fiber node. A reasonable MPEG2 delivery to set-top-boxes. IPTV is by endgame scenario for HSD would provide definition an IP packet with video content 100 Mbps channel bonded service to 50% of from an IP source to an IP destination address. households passed with a 0.25% concurrency, In traditional IPTV, an IPTV server sends for a total HSD fiber node throughput an IP packet to the destination address of a requirement of 750 * 0.5 * .0025 * 100 Mbps subscriber playback device, which we call = 94 Mbps. Thus, under reasonable generically an “IP Set-Top Box” (IPSTB). hypothetical endgame scenarios, VOD is With the M-CMTS architecture forwarding, expected to require 26 times the bandwidth of the IPTV packet is routed through the M- HSD. CMTS core. In a conventional M-CMTS A key goal of the DOCSIS Modular architecture without DIBA, therefore, the CMTS (M-CMTS) architecture was to define IPTV content is forced to make two transits a common Edge QAM (EQAM) architecture through the Converged Interconnect Network that could support a cost-effective transition (CIN) switch that connects regional networks from Digital Video (DV) VOD delivery to with the core CMTS. IPTV VOD delivery over DOCSIS. A The video/IP traffic travels to the M- straightforward strategy of using M-CMTS to CMTS core, then “hairpins” back through the provide IPTV VOD would call for deploying CIN on a DOCSIS External Physical Interface M-CMTS EQAMs for DV-VOD first, and (DEPI) pseudo-wire to the DEPI EQAM. If- adding M-CMTS core capacity as needed to and-when IPTV becomes a significant transition the DV-VOD to IPTV-VOD. As we fraction of overall bandwidth delivered to a have seen, though, the M-CMTS core fiber node, this hairpin forwarding of IPTV capacity used for IPTV-VOD will quickly content will require significant expenditures exceed the CMTS capacity used for HSD. A by MSOs for M-CMTS core and CIN straightforward implementation of M-CMTS switching bandwidth. for IPTV, therefore, will require many times In DIBA, rather than pass through the M- the cost of M-CMTS core capacity as that CMTS core, the high-bandwidth video/IP currently used for HSD. content is tunneled from the IPTV server,

through the MSO’s converged interconnect bypass traffic and omit these functions. With network directly to the DIBA EQAMs. The an integrated CMTS and no timestamps in the IPTV emerges from the DIBA EQAM with un-synchronized channels, the DOCSIS full DOCSIS framing, suitable for forwarding Timing Interface which is required in the M- through a DOCSIS cable to home IP CMTS architecture is not necessary in DIBA. devices. This architecture will deliver high- In DIBA, a DOCSIS 3.0 bandwidth entertainment video/IP/DOCSIS to requires only one synchronized channel from the home for a price-per-QAM that matches the existing integrated CMTS to provide that of the most advanced, cost reduced timing, control the upstream transmissions, MPEG2 EQAMs. and provide the other MAC functions. The DOCSIS 3.0 cable modem will receive REDUCING DELIVERY COSTS video/IP on additional un-synchronized, inexpensive, DIBA-generated channels, which Operators can deploy independently can even be bonded. scalable numbers of downstream channels without changing the MAC domain or the LEVERAGING EXISTING EQAMS number of upstream DOCSIS channels. These downstream channels are available for DIBA offers a variety of bypass VOD/IP and switched-digital-video/IP. They encapsulations. Using a standard M-CMTS can also lower the cost to deliver video over DEPI EQAM, two bypass encapsulations are DOCSIS service to be competitive with possible, depending on the video server and today’s MPEG VOD. DIBA accomplishes the MSO network. In either case, the server goals of the M-CMTS without the originates a Layer 2 Tunnel Protocol Version unnecessary expense of the M-CMTS. DIBA 3 (L2TPv3) tunnel to the DEPI EQAM. The avoids the expense of the DOCSIS MAC tunnel payload can be: domain technology for the video/IP traffic by using both synchronized and unsynchronized DOCSIS downstream channels. The • The DOCSIS Packet Streaming synchronized channels pass through the Protocol (PSP) in which the video/IP integrated CMTS or the CMTS core and is encapsulated into DOCSIS MAC provide the many DOCSIS MAC functions, frames. This permits the EQAM to including: mix both IPTV traffic originated from the IPTV server with non-IPTV (e.g. Conveying the DOCSIS timestam • ps HSD or VOIP) traffic originated from • Managing ranging to provide the the M-CMTS core on the same proper time-base to the cable modem DOCSIS downstream carrier. • Instructing the cable when to transmit upstream • The DOCSIS MPEG Transport (D- MPT) layer which consists of an • Delivering other MAC layer messages MPEG-2 Transport Stream of 188 byte for cable modem registration, packets. All DOCSIS frames, maintenance, etc. including packet-based frames and In contrast, the unsynchronized DOCSIS MAC management-based frames, are channels are generated by the EQAMs included within the one D-MPT . (including installed MPEG EQAMs) for the The EQAM searches the D-MPT payload for any DOCSIS SYNC Switched broadcast IPTV multicast is sent messages and performs SYNC without BPI on non-synchronized corrections. It then forwards the D- downstream channels to DOCSIS 3.0 cable MPT packet to the RF interface. The modem IPTV devices. intent of D-MPT mode is to allow MPEG packets to be received by the THE ROLE OF THE CONTROL PLANE EQAM and forwarded directly to the RF interface without having to While there is no design yet for the DIBA terminate and regenerate the MPEG control plane, there are several other control framing. The only manipulation of the planes that must be considered in that design. D-MPT payload is the SYNC The current MSO switched broadcast correction. architecture is intended to save bandwidth for A standard MPEG2-Transport Stream a moderate-sized optical node, and that (MPEG2-TS) EQAM can also be used. Here feature will have to be retained in any DIBA the video server may transmit an IPTV PSP implementation. The PacketCable Multimedia formatted data packet. A PSP/MPT converter, (PCMM) architecture is intended to provide either attached to or embedded within the CIN QoS within a DOCSIS system, and that networking device, then changes the format function will have to be retained. The M- into an MPEG-2-TS that a conventional CMTS architecture is intended to provide a MPEG EQAM can process. An alternative is scalable number of downstream DOCSIS for the VOD server to directly generate channels managed by an edge resource IPTV/MPT formatted packets that the MPEG manager, and this feature is an essential part EQAM can process. In the case of non- of DIBA. IP Multimedia Subsystem (IMS)- synchronized DOCSIS channels, a non- based IP services will include entertainment DOCSIS Program ID (PID) is used because video as well as personalized and on-demand each D-MPT program would require a video. separate MPEG-2 PID. A required extension for sending multiple programs streams of D- VOD AND SWITCHED BROADCAST MPT to the same downstream QAM channel would be for DOCSIS 3.0 cable modems to be Many cable operators have existing VOD programmed to accept D-MPT-formatted and switched digital video architectures. packets with other than the standard DOCSIS These two architectures have quite different PID. purposes, but are similar in one major way to conventional digital broadcast video. In all TUNNELING BROADCAST IPTV cases the video is brought over the MSO network to the EQAMs as MPEG-2 There are several tunneling options that SPTS/UDP/IP. The EQAM then strips off the can be used. One of these is “Interception”, in UDP/IP encapsulation, multiplexes the which SPTS/UDP/IP encapsulation is material into multi-program MEPG-2 TSs that intercepted by a CIN multicast router and are transmitted over QAM carriers. The set- encapsulated into PSP tunnel. Another is PSP top boxes are able to demodulate the MPEG-2 Tunnels directly from the video server to the TSs and decode the video. Since the set-tops DEPI EQAMs. For bonded IP multicast, a cannot receive video over an IP stack, they “DIBA Distributor” component co-located must be instructed which QAM carrier with a CIN IP multicast router distributes or frequency and which Program ID to “stripes” the IP packet to a pre-configured demodulate. In the case of switched digital DOCSIS 3.0 Downstream Bonding Group.

video, the objective is to make more video for IPTV. In one, the client (video viewing) channels available than there is bandwidth for device does not itself support the QoS in a typical RF spectrum. All the available signaling mechanisms and relies on the video content is carried as a collection of IP Application Server as a proxy. The client multicast sessions, in most cases with all sends a service request, and the Application content available within the hub. Only when a Server sends a service request to the set-top within a fiber node requests a Application Manager. Upon receipt of this particular video title is that title carried to that request, the Application Manager determines node. the QoS needs of the requested service and Statistically, some number of set-tops sends a Policy Request to the Policy Server. will be viewing each of the popular video The Policy Server in turn validates the Policy broadcast titles. Other, less popular titles will Request against the MSO-defined policy rules likely be viewed by only one set-top at a time. and, if the decision is affirmative, sends a Still other less popular titles will not be Policy Set message to the CMTS. The CMTS viewed at all. Thus, the limited bandwidth performs admission control on the requested available to that node will accommodate a QoS envelope (verifying that adequate great many more available titles than could be resources are available to satisfy this request), carried simultaneously. installs the policy decision, and establishes the service flow(s) with the requested QoS When the control plane receives a request level. from a set-top for a new title, it searches for an EQAM serving that node which has sufficient available bandwidth. This EQAM is M-CMTS EDGE RESOURCE MANAGER instructed to issue an IGMPv3 join to that IP multicast session. In the case of VOD, the The M-CMTS architecture takes the control plane must locate a server with the PCMM architecture one step further in terms desired title. This server will then send the of granularity. Within the PCMM title as unicast MPEG2-TS/UDP/IP to an architecture, it is the CMTS that makes the appropriate EQAM serving the set-top making admission control decision for new flows and the request. In the case of VOD, there can be assigns then to a suitable DOCSIS channel. In thousands of available titles. the M-CMTS, the core CMTS must in turn request QAM resources from the edge PACKETCABLE MULTIMEDIA resource manager.

PacketCable Multimedia will be an IP VIDEO AND IMS important element of the control plane for DIBA. The objective of managing QoS for the There has been much work within the myriad of services to be provided over the Internet community to use IMS to control limited bandwidth of DOCSIS 1.0 and 2.0 video services over IP to offer roaming and networks led to the development of PCMM. two-way features. This has generally not been PCMM uses such elements as the Application for entertainment video. However, the current Server, Application Manager, Policy Server, aim of IMS has become the universal and CMTS. These are all in the headend and deployment of all IP-based services through under control of the MSO. PCMM is only both fixed and mobile networks, regardless of available for IP-based services. There are location. So it is possible that IMS will several scenarios for the operation of PCMM become a suitable vehicle for the deployment of entertainment video as well, particularly application servers to route particular personalized and on-demand video. messages to. This is an aggressive goal, but not an unreasonable one, given the history of and 3. Interrogating-CSCF is another SIP effort that has gone into IMS. IMS began as proxy, this one serving as the interface an effort by telecom carriers in the Third by which IMS messages from the Generation Partnership Project (3GPP) to outside world access the local converge mobile services with VoIP and IP network. The I-CSCF is able to inspect multimedia by using the Session Initiation the user database to determine, for a Protocol (SIP). SIP is an IP-based peer-to- particular IMS terminal, the correct peer protocol used to establish and control Server-CSCF to forward signaling two-way flows carrying voice, video, and messages. gaming. The actual flows use Real-Time Protocol (RTP) and UDP/TCP. True to its heritage, non-IP devices such as analog The network also has application servers telephones are supported through gateways. that are linked to the control plane via the Server-CSCF. In the past, these applications Gradually the range of services and have been call related, such as call-waiting, access technologies increased, resulting in a call-forwarding, conference calls, voicemail, powerful Next Generation Networking (NGN) etc. However, it is possible to have these architecture. The access technologies include services include such network-oriented tasks any IP/SIP-enabled devices, including phones, as resource management for VOD sessions. PDAs, computers over DSL, DOCSIS, 802.11, etc. The network control plane is built There is an obvious advantage to using around a Call Server Control Function IMS and SIP to provide entertainment video (CSCF) that comprises the following: services in that the control plane would naturally extend to any IP video sources in the Internet on the one hand, and to a multitude of 1. Proxy-CSCF, which is an initial fixed and wireless video playback devices on control plane element to interact with the other hand. The roaming ability and the an IMS device and acts as a proxy in device presence enable handoff between the SIP processing. Beginning with devices, a necessary component of seamless registration, each IMS terminal is mobility. There are many issues to deal with assigned to a Proxy-CSCF. The Proxy- in integrating the VOD and switched digital CSCF authenticates the IMS device video control and data planes with IMS/SIP. and establishes security. It also authorizes the IMS device to use BROADCAST IPTV OVER DIBA network bandwidth, and may apply QoS policies to this use. It is in the path of all signaling. In the typical MSO approach to switched broadcast video, the set-tops are non-IP so it 2. Server-CSCF, which is a SIP server is the MPEG EQAM that joins the IP and binds the user IP address to its SIP multicast group. The multicast group is then address. It interfaces to the user mapped to a particular QAM and PID, and it database (the Home Subscriber Server is this information that is forwarded to the set- or HSS) where user information is top to enable it to receive the MEPG-2 video kept and sees all signaling messages. It transport stream. also determines which of the

In fact, DOCSIS is not generally used to deliver broadcast entertainment video, since • Data Plane operation from IPTV server the bandwidth is too expensive. However, the to DIBA PSP/MPT Converter, DIBA intention of DIBA is to bring the cost of Distributor, and DEPI/MPEQ EQAMs DOCSIS bandwidth low enough to make

IPTV viable. With DIBA, the IP set-top will be able to join the IP multicast group. The • NGOD Session Manager control switched broadcast control plane determines signalling to IPTV server which programs are multicast to which fiber node, consistent with the IP set-tops receiving • CMTS control signaling to DIBA which programs. Distributor and PSP Converter functions in CIN BROADCAST IPTV TO DOCSIS 2.0 AND 3.0 MODEMS • CMTS control signaling to EQAM Broadcast IPTV is possible with both DOCSIS 2.0 and 3.0 modems. In either case, DIBA will remove the boundaries of the IP set-top box sends an IGMP join to an delivering video using a M-CMTS, and it will IP multicast session for a particular program. allow cable operators to leverage service The DOCSIS 2.0 modem will have to use a velocity by efficiently delivering IPTV synchronized DOCSIS downstream channel, services that bypass the M-CMTS core. while a DOCSIS 3.0 modem can receive on a Bonded Channel Set. It may or may not be ABOUT THE AUTHOR necessary to change DOCSIS channels to change to a new IPTV ‘channel’. However, Mike Patrick is a Data Networking any DOCSIS channel changing requires Architect, Motorola Connected Home communication from the CMTS to the cable Solutions. In this role, He is responsible for modem in the form of a Dynamic Channel the architecture of Motorola’s Cable Modem Change (DOCSIS 2.0) or Dynamic Bonding Termination System products (CMTS). He is Change (DOCSIS 3.0) to change a tuner of currently leading technical development of the the cable modem to the new channel. DOCSIS® 3.0 feature set for Motorola’s BSR 64000 CMTS/Edge Router. SUMMARY Mike joined Motorola in 1994. Over the last 13 years, Patrick was a Data Networking Cable operators will be able to System Engineer and was a major contributor economically supply IPTV capacity of to the Downstream Channel Bonding and IP gigabits per fiber node by bypassing the M- Multicast features of DOCSIS 3.0. Mike was CMTS core and tunneling IPTV directly to a major contributor to version 1.1 of the Data the M-CMTS EQAM. DIBA allows operators Over Cable System Interface Specification to avoid the cost of additional M-CMTS core (DOCSIS), which standardized QoS operation capacity for IPTV. DIBA is proposed as a for VoIP cable modems. work item for CableLabs after DOCSIS 3.0 Before joining Motorola, Mike served in draft specifications. If it is approved, various engineering and management CableLabs would standardize: positions at Digital Equipment Corporation, General Computer Corporation, Cryptall Corporation, Ztel, and . He has 14 U.S. patents granted to date. Mike earned his bachelor’s of science and master’s of science, computer science degree from the Massachusetts Institute of Technology in 1980, as well as a Masters in the Management of Technology from MIT Sloan School in 1992. He can be reached at [email protected].