Blueprint: Network Architecture

DOCSIS 3.0, Raising the Standard

by Ran Senderovitz and Etai Zaltsman, Broadband Technology Group

11/26/2007

The newly enhanced version of the cable industry's data communications protocol, DOCSIS 3.0 (Data Over Cable Service Interface Specification), ushers in an era of breathtakingly faster line speeds supporting new applications that hold the promise of expanded revenue streams for operators. And, a cost-effective migration path means multi-service operators (MSO) can deploy DOCSIS 3.0 sooner and provision it later as marketplace demand warrants the roll-out of new services.

At a minimum, DOCSIS 3.0 starts by increasing a subscriber's bandwidth up to 160 megabits per second (Mbps) downstream and 120 Mbps upstream. That's significant bandwidth and certainly adequate to handle a slew of high definition (HD) multimedia applications. In addition, more and more cable operators are generating sizeable revenues from voice-over-cable (VoCable) services that bring them into direct competition with telecomm companies. There is also Internet Protocol (IP) television (IPTV) on the horizon, which will eventually become a more flexible, efficient and scalable methodology for television programming distribution. All of these services play into the hands of the next-generation capabilities of DOCSIS 3.0.

Some Bonding Time

Enabling the higher line speeds of DOCSIS 3.0 is the concept of channel bonding. Essentially, 3.0 builds on the channel structure of its predecessor specification, DOCSIS 2.0. Taking the DOCSIS 2.0 channel structure as a starting point was critical for DOCSIS 3.0 because this ensures backward compatibility with legacy equipment and, moving forward, it enables heterogeneous networks where both 2.0 and 3.0 cable modems interoperate transparently and efficiently.

DOCSIS 3.0 combines or bonds together the physical 6 MHz channels of DOCSIS 2.0 to achieve higher bandwidth logical channels. DOCSIS 3.0 takes four or more 6 MHz DOCSIS 2.0 physical channels rated at 40 Mbps (or the 8 MHz channels rated at 50 Mbps that are the norm in Europe) and binds them together as one logical channel. The bound channels need not be contiguous or have the same performance parameters, such as line speeds, latencies and others.

The minimum level of channel bonding a DOCSIS3.0 modem should support begins with four channels, but any number of channels can be logically bound together to deploy much higher bandwidth where it is needed. So, businesses, for example, could subscribe to a DOCSIS 3.0 service that is based on eight bound channels and delivers a bandwidth pipe capable of data transfer speeds of 320 Mbps downstream. Bonding 16 channels together would create a pipe with a downstream bandwidth of 640 Mbps. The flexibility and scalability inherent in the concept of channel bonding gives operators a high degree of precision when it comes to efficiently deploying and provisioning bandwidth capacity to the exact locations where it is needed in their networks.

Making it Work

In the downstream mode, DOCSIS 3.0's channel bonding is achieved when a network processing node, such as a cable modem termination system (CMTS), transmits legacy DOCSIS 2.0 packets on multiple physical channels that are logically bonded together. (Figure 1) Unlike DOCSIS 2.0's downstream data transmissions which are essentially a continuous transmission of an MPEG transport stream over one channel, a DOCSIS 3.0 packet stream arrives at a cable modem on one logical channel made up of multiple physical channels. And since each physical channel can have different performance parameters, the packets can arrive out of sequence. For example, packet number 1 on a slower physical channel might arrive after packet 2 arrives because packet 2 was delivered via a speedier physical channel. It is up to the modem to re-assemble the packets from the multiple channels into one packet stream.

To accomplish this re-assembly process, the DOCSIS 3.0 specification has added several new fields to the DOCSIS 2.0 packet header. These new management fields prioritize the packet for quality-of-service (QoS) considerations, establish a sequence number and define a Downstream Service Identification (DSID) for the packet. By examining each packet's sequence number, the cable modem ascertains the correct order of the packets in a stream and re-orders those that have been delivered out of sequence.

Of course, latencies can develop when packets arrive at the cable modem out of sequence. A late-arriving packet could delay the re-assembly process accordingly. DOCSIS 3.0 defines a maximum latency of 18 milliseconds (ms) during which time the modem waits to receive a packet. After 18 ms the packet is declared lost if it has not yet arrived.

Some sophisticated modem platforms have introduced algorithms that can reduce this latency even further, ensuring higher service quality for time-critical applications such as VoCable, IPTV and others. One such algorithm declares a packet lost as soon as subsequent packets have been received on all of the bonded channels, which can often occur before the 18 ms has elapsed. Reducing latencies can be a critical cost factor for cable modem architectures because higher latencies require more memory in the modem to buffer packets during the re-assembly process.

Latencies that might develop in the network can be avoided by judiciously assigning the channels within the bonded bundle of channels to certain applications. This is accomplished through the DSID header field. For example, a time-critical application like IPTV or VoCable might be assigned through DISD header information to the two channels in a three-channel bundle that share identical performance characteristics, including line speeds. This would ensure that packets arrive at the modem very close to their correct sequence and latencies would be kept to a bare minimum. Or, VoCable service, for instance, could be assigned to just one channel, eliminating any out-of-sequence packets and network latencies from channel bonding. Conversely, an application that is not time-critical and only requires a "best effort" type of service, such as Internet access, might be assigned through the DSID field information to all of the channels in a bundle with little regard for the miniscule latencies that might develop.

For upstream communications from the cable modem to processing nodes in the network such as CMTSs, the new channel bonding techniques in DOCSIS 3.0 have eliminated some of the inefficiencies inherent in DOCSIS 2.0. For example, DOCSIS 3.0 introduced the concepts of continuous concatenation and fragmentation (CCF). In both DOCSIS 3.0 and 2.0, CMTSs can allocate time slots (called mini-slots in DOCSIS language) in the network to accommodate a cable modem's request for supplemental bandwidth. In DOCSIS 2.0, the modem is restricted to requesting time slots in increments adequate for transmitting entire packets. So if a modem has a packet of 100 bytes in an upstream communication queue, DOCSIS 2.0 requires that the modem can not request additional bandwidth until it is granted the entire 100-byte packet. DOCSIS 3.0 loosens some of these rigid restrictions so that data can be transmitted as soon as a reasonable amount of bandwidth is available. With DOCSIS 3.0's continuous fragmentation feature, part of the data in a packet in a modem's communications queue can be transmitted immediately if the bandwidth is available and allotted by the CMTS. The remaining data will be sent later when additional bandwidth is freed up. With continuous concatenation, the data from more than one packet can be combined in one transmission to improve bandwidth efficiencies. (Figure 2)

DOCSIS 3.0 includes a number of other concepts such as upstream service group IDs and Service ID (SID) clusters that also facilitate logical channel bonding and higher line speeds. But, in addition to its pure performance-oriented aspects, DOCSIS 3.0 has several deployment and operational features that will make a cable operator's migration to this technology much more efficient and revenue-effective.

Deploying Sooner Rather Than Later

Although a leap forward in capabilities, the DOCSIS 3.0 specification remains firmly rooted in DOCSIS 2.0 and other legacy technologies. This is a good thing for cable operators because it enhances the ease and cost-effectiveness with which DOCSIS 3.0 modems can be deployed today even if their advanced capabilities may not be provisioned for some time. In fact, deploying DOCSIS 3.0 modems today can pay off handsomely later when new multimedia and other high-bandwidth services can be rolled out quickly without replacing all of the end user modems that have already been installed.

Since DOCSIS 3.0 and DOCSIS 2.0 modems are interoperable, a cable network could evolve as a heterogeneous environment with both types of modems in the field. Operating expenses are not impacted significantly by this as the network's management system will be able to transparently interact with and control both types of modems as if they were identical.

With little or no effect on operating costs, the timing of a MSO's migration to DOCSIS 3.0 may come down to procurement costs for cable modems. Thanks to some innovative architectures and newly developed technologies, the cost differential between DOCSIS 3.0 and DOCSIS 2.0 modems can be kept to a minimum. For example, instead of implementing several individual tuners for each of the multiple physical channels in a bonded channel bundle, TI's Puma 5 DOCSIS 3.0 modem platform keeps its bill of material (BOM) costs low by featuring one innovative wideband tuner that can handle as many as eight physical channels, reducing significantly the cost that discrete tuners would generate.

If DOCSIS 3.0 modems were to begin being deployed today, the advanced aspects of DOCSIS 3.0 would already be in place when the market or even particular users demanded new services. For example, the fact that DOCSIS 3.0 supports the expanded addressing scheme of Internet Protocol Version 6 (IPv6) could eventually simplify the roll-out of VoCable services. It is expected that the transition from DOCSIS 2.0 to DOCSIS 3.0 for VoCable services will be seamless.

Another capability that has been significantly enhanced in DOCSIS 3.0, advanced multicasting, will be very useful to operators when the time comes to migrate to IPTV. Of course, IP bandwidth is critical to an IPTV system, and multicasting conserves bandwidth by providing it much more efficiently in applications with the characteristics of IPTV. DOCSIS 3.0's advanced multicasting capabilities readily identify multiple users who are receiving the same packet stream and place them in a multicasting group. In applications like IPTV where a group of subscribers may be viewing the same content, this can be an efficient way of allocating network bandwidth. The network management system can dynamically add to or remove subscribers from a group and transmit the same packet stream once to all subscribers in the group. Previously under DOCSIS 2.0, packet streams were individually transmitted to each user, consuming significantly more network bandwidth and management resources.

The flexibility and scalability of the DOCSIS 3.0 cable modem deployed could also facilitate an orderly transition to IPTV in the future. For example, a cable modem with four channel receivers or a single wideband receiver such as the one implemented in TI's cable modem architecture could support one high-speed DOCSIS data channel and legacy video channels. Over time, a cable operator might make a strategic decision to shift to IPTV for its more effective management capabilities. In this case, the cable modem could be dynamically re-configured, designating additional channels as DOCSIS data channels for IPTV. Or, the cable modem could evolve into a DOCSIS 3.0 set-top gateway where IP video is distributed over a home IP network to various IP set-top-boxes (STB) throughout the residence.

Marketplace Potential

DOCSIS 3.0 holds great potential. But unlike other technological breakthroughs that may have been ahead of the marketplace, cable operators are able to take advantage of DOCSIS 3.0's potential today while the market is still developing. The best time to take advantage of that potential is now. Doing so will have even greater payoff in the future when applications needing DOCSIS 3.0's advanced capabilities become commonplace. Tomorrow's and many of today's new applications that are multimedia rich and offer compelling and engaging functionality will quickly absorb the elevated bandwidth of DOCSIS 3.0, much to the delight of subscribers.

Of course the timing of a DOCSIS 3.0 deployment will vary from operator to operator. Some may wait and watch while the more aggressive MSOs begin implementing DOCSIS 3.0 modems in their networks. These more aggressive operators will be poised to capitalize on burgeoning subscriber demand for new data-centric applications. Those who wait and watch the market develop may find themselves playing catch-up in a competitive marketplace with an increasing number of alternatives available to consumers.

About the Authors

Etai Zaltsman is the chief technical officer of the Broadband Technology Group at Texas Instruments and Distinguished Member of the Technical Staff. Zaltsman leads TI's Herzelia site and works to define long term roadmaps and vision for his group while driving next generation architectures, technology leadership and innovation throughout his site.

Zaltsman joined TI 11 years ago as the chief architect and product marketing manager as he led the architecture and development of a DOCSIS cable modem chipset and systems in the company. He lists the key milestones at his current position as the driving of the DOCSIS 3.0 standard and industry, defining PUMA5 architecture, defining cable video strategy and initiating home networking activity.

Zaltsman received his Bachelor of Science in Electrical Engineering in 1998 and his Master of Science in Electrical Engineering in 2004, both from the University of Tel Aviv. His published thesis is titled Methods for Improving the Performance of Digital Communication Systems by Using Soft Decision Decoding. Currently, he is in the last year of study for executive MBA at Kellogg-Recanati, a joint program of Tel Aviv University and Northwestern.

Ran Senderovitz, cable business manager of Texas Instruments' (TI) Cable Modem Business Unit, is responsible for driving business strategy and planning the business' roadmap for profitability. Under Senderovitz's direction, TI now retains the leadership position in EMTA DOCSIS 2.0 technology and continues to drive Puma 5 and DOCSIS 3.0 execution.

Senderovitz started his TI career in 1999 as a system test manager, and soon became the system development manager for the cable modem and platform support packages. In 2003, Senderovitz transitioned to the WLAN business unit as a system and software development manager, and then system architecture and advanced technology manager. Prior to his current position as general manager, Senderovitz served as the engineering director of the cable modem business unit responsible for managing the engineering activates at TI for the cable modem market.

Senderovitz began his career with the R&D Unit in the Israeli Defense Force, where he spent eight years and was released at the rank of a Major. He received his bachelor of science in electrical engineering from Tehcnion, Haifa in 1992 and his master of science in electrical engineering from Tel-Aviv University in 2002. Additionally, in 2007 he earned his executive MBA from Kellogg-Recanati (a joint program of Tel Aviv University and Northwestern).

About Texas Instruments

Texas Instruments Incorporated provides innovative DSP and analog technologies to meet our customers' real world signal processing requirements. In addition to Semiconductor, the company's businesses include Sensors & Controls, and Educational & Productivity Solutions. TI is headquartered in Dallas, Texas, and has manufacturing, design or sales operations in more than 25 countries.

Send us your response to this article.

Learn How to Get Your Column Published on this Site