◆
Next-Generation Access Networks: A Preview
Marcus K. Weldon, Thierry Van Landegem, and Edward S. Szurkowski As we enter 2008, fixed line access networks are in a full state of flux or “transformation,” across many dimensions. For example, central office asymmetric digital subscriber line 2 (ADSL2)-centric copper networks are rapidly evolving to remote node-based very high speed digital subscriber line 2 (VDSL2)-based deployments as well as to direct fiber to the home (FTTH) networks, with an accompanying shift from asynchronous transfer mode (ATM)-based to native Ethernet-based transport. Coupled to these shifts is a manifest evolution in the services being provided, with voice service evolving from circuit-switched plain old telephone service (POTS) to consumer Voice over Internet Protocol (VoIP), data services evolving from highly asymmetric bandwidth services with downstream data rates of a few megabits per second to more symmetric bandwidth with 30 Mbps to 100 Mbps data rates, and a full array of video services supported. Furthermore, there is an increasing expectation on the part of the end user that these services will be easily accessible on any device, on any network, at any time, with a guaranteed quality of experience (QoE). In addition to a further evolution in the network infrastructure, these extended requirements give rise to the need for a next generation of network and service management, with dynamic management of the access network resources, sophisticated service-level diagnostics, as well as management of the network inside the home. This issue of the Bell Labs Technical Journal is dedicated to a discussion of these topics and the requirements they put on the access network in the near and long term. © 2008 Alcatel-Lucent.
The First Phase of Network Transformation: The Move to “All IP” Over the past few years, the traditional telecom operator’s network has undergone a transformation to an “all Internet Protocol (IP)” network, with the amalgam of legacy networks that had been built to provide residential and business voice and data services being replaced by a single packet-based IP/Ethernet and IP/multiprotocol label switching (MPLS) network. One of the main drivers of this transformation was
the lower operational expenditure (OPEX) associated with operating and maintaining a single network, with typical operator estimates for the magnitude of these savings in OPEX being on the order of $100 per subscriber per year. In addition to the apparent cost savings associated with the transformation to an “all IP” network, the intent was to increase the average revenue per user (ARPU) by offering a full complement
Bell Labs Technical Journal 13(1), 1–10 (2008) © 2008 Alcatel-Lucent. Published by Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com) • DOI: 10.1002/bltj.20278
Panel 1. Abbreviations, Acronyms, and Terms ADSL2—Asymmetric digital subscriber line with extended rate and reach characteristics AN—Access node ARPU—Average revenue per user ATM—Asynchronous transfer mode CPE—Customer premises equipment DHCP—Dynamic Host Configuration Protocol DVR—Digital video recorder FTTN—Fiber to the node FTTH—Fiber to the home GPON—Gigabit passive optical network HD—High definition HSI—High speed Internet HTTP—Hypertext Transfer Protocol IMS—IP Multimedia Subsystem IP—Internet Protocol IPTV—IP television ISDN—Integrated services digital network LAN—Local area network MAC—Medium access control MPEG—Motion Picture Experts Group MPLS—Multiprotocol label switching NAT—Network address translation NGN—Next-generation network nPVR—Network PVR
of video services, including standard definition (SD) and high definition (HD) broadcast television (TV), as well as video on demand (VoD) and personal video recorder (PVR) services. The value of such services varies greatly by region and country, but typically falls in the range of €20 per subscriber for a basic package of services in Europe, to $100 or more per subscriber in North America. The elements of this first phase of transformation are shown in Figure 1 for data or “high speed Internet” (HSI) services, highlighting the shift from session-based Point-to-Point Protocol (PPP) to connectionless Dynamic Host Configuration Protocol (DHCP)-based access, with native IP/Ethernet transport in the access layer, and IP/MPLS-based aggregation and core networks that support high availability, resiliency, and security with per-subscriber, per-service hierarchical quality of service (QoS). In the new architecture, the network, service, and subscriber management is centralized, with the services intelligence
2
Bell Labs Technical Journal
DOI: 10.1002/bltj
OPEX—Operational expenditure P2P—Peer-to-peer PC—Personal computer PON—Passive optical network POTS—Plain old telephone service PPP—Point-to-Point Protocol PVR—Personal video recorder QoE—Quality of experience QoS—Quality of service RAC—Resource admission control RG—Residential gateway S/BC—Session border controller SD—Standard definition SDP—Service delivery platform SIP—Session Initiation Protocol STB—Set-top box TISPAN—Telecommunication and Internet converged Services and Protocols for Advanced Networks TV—Television VDSL2—Very high speed digital subscriber line 2 VLAN—Virtual LAN VoD—Video on demand VoIP—Voice over Internet Protocol
becoming distributed among the access, aggregation, and core elements to ensure multi-dimension services scalability. The combination of these functional elements not only allowed QoS-enabled data services, but, critically, also allowed full support of voice and video services—the so-called triple play of services— over the same network infrastructure. In order to fully support triple play services, in addition to the functional architecture transformation shown in Figure 1, a wholesale increase in the aggregate end per-user bandwidth was also required. Referring to Figure 2, in the access layer, the bandwidth required to support the required array of standard and high definition video services, along with voice and data services, is around 25 Mbps downstream, and approximately 1 Mbps upstream. This bandwidth asymmetry in the access portion of the network is a reflection of the persistence of professionally generated video programming and web content that is sourced from a centralized location to the
ATM aggregation network
Legacy infrastructure ATM switch
DSL gateway 1
IP/fiber in access (VDSL2, GPON)
Traditional BRAS
ADM 3
2
Session-based to connectionless
Transport network
ATM DSLAM
5
4
ATM to Eth/MPLS aggregation
Internet
TDM to packet transport
BE to QoS/HA non-blocking
Cost optimization and subscriber control All IP architecture Triple play
BSAN
BSR BSA
Residential gateway
Service reach
MSPP/WDM
Plug and play service flexibility ADM—Add/drop multiplexer ATM—Asynchronous transfer mode BE—Best effort BRAS—Broadband remote access server BSA—Broadband service aggregator BSAN—Broadband service access node BSR—Broadband service router DSL—Digital subscriber line DSLAM—Digital subscriber line access multiplexer
Service quality reliability
Bandwidth GPON—Gigabit passive optical network HA—High Availability IP—Internet Protocol MPLS—Multiprotocol label switching MSPP—Multiservice provisioning platform QoS—Quality of service VDSL2—Very high speed digital subscriber line 2 WDM—Wavelength division multiplexing
Figure 1. The essential elements of the first phase of network transformation.
user, with the user only generating a limited amount of upstream traffic in the form of control commands (e.g., channel changes, movie orders, and Hypertext Transfer Protocol [HTTP] requests). Notably, access network technologies have been designed to take advantage of this inherent services asymmetry, in order to minimize the cost per user, by utilizing less spectrum and lower-order modulation schemes in the upstream than in the downstream direction. In the aggregation and core transport network, the required bandwidth is seen to be on the order of
100 Gbps, with a much lower level of statistical multiplexing possible, due to the higher persistent bandwidth requirements that result from providing “always on” video services. For example, a multiplex of 500 MPEG-2 encoded HDTV channels or, equivalently, multiple redundant copies of SDTV channels, requires approximately 10 Gbps of bandwidth. If this is combined with support for one dedicated 5 Mbps SD “on demand” stream per user, with 100,000 users and 20 percent services concurrency, bandwidths on the order of 100 Gbps are required, as indicated. Notably,
DOI: 10.1002/bltj
Bell Labs Technical Journal
3
More bandwidth to the end user
Less oversubscription in the network Nominal bandwidth
SD channel – 2.3 Mb/s
20
HSI – Mb/s to 5 Mb/s (3 Mb/s)
15 Mbps
HD channel – 9.3 Mb/s SD channel – 2.3 Mb/s
10 5
SD channel – 2.3 Mb/s VoIP – 160 Kb/s
0
Voice/video 85% traffic volume and revenue contribution
Oversubscription Over subscription ratio
Estimated peak bandwidth levels - 2010
HSI
70 60 50 40 30 20 10 0
IPTV
100 Gbps 0.5 Tbps (symmetric)
25
Access network
80 Mbps
Subscriber/service aggregation
HSI
IPTV
Long haul network Service edge
~8,000 users 1
HD—High definition HSI—High speed Internet IP—Internet Protocol
30 Mbps
~300 users
~80,000 users
SD—Standard definition TV—Television VoIP—Voice over IP
Figure 2. The typical bandwidth requirements for a “triple play” network.
the same services asymmetry is also generally apparent in the aggregation and core networks as was highlighted above; however due to the lower cost sensitivity in these layers of the network and the exclusive use of optical transmission (which does not suffer from the same interference and signal attenuation issues that limit copper-based transmission), the capacity in the aggregation and core layers is typically symmetric in nature.
The Second Phase of Network Transformation: Intelligence Moves to the Edge The distribution of services intelligence that was described above as one of the underpinnings of the first phase of network transformation is a trend that continues today as we approach the second phase of transformation, which is outlined below, and throughout this issue. But before describing this further
4
Bell Labs Technical Journal
DOI: 10.1002/bltj
network evolution, it is important to recognize the changes in the end-user services that are driving further transformation. Next-Generation Services and Bandwidth Requirements Referring to Figure 3, the most fundamental shift in the services model is that the concept of shared bandwidth, in the form of either statistical multiplexing or multicasting, is being replaced by dedicated per-user bandwidth, with high levels of concurrency (simultaneous usage by all users). In addition, the requested content is no longer generated by a relatively small number of nationally based professional sources (e.g., studios) and sourced from a centralized point (e.g., a video head-end), but is generated by a global array of user-producers and consumed on the users’ timescale and on the device of their choosing. The latter requirement—ubiquitous availability—leads directly to the
Multicast
Unicast
Network schedule Limited content library Content streamed to users Home storage
User schedule “Internet of content”
Content streamed from users Network(ed) storage
Single viewing device (TV) Single, fixed location
Any device
Any location
TV—Television
Figure 3. Services evolution driving the continued evolution of the network.
need for such functions as network-based caching and media transcoding in order to provide seamless, rapid access to content at the optimal bit rate and format. One other significant evolution in the services model is that individual voice, data, and video services that previously only interworked in the most rudimentary fashion, e.g., a telephony caller ID appearing on a TV screen in the IPTV “domain,” will be more tightly integrated to allow one service to enhance or facilitate another. One such example might be using a mobile phone (traditionally a telephony and messaging device) to control devices in the home network, or to exchange related media content with a PVR in an IP television (IPTV) set-top box (STB), using a combination of a residential gateway and a femtocell as mediating devices. Such a service would likely involve interworking of applications within the home network. Other service “blends,” such as the discovery and display of Internet TV content (e.g., NBC.com or YouTube* content), alongside the conventional broadcast IPTV content, or roaming on video content between wireline and wireless networks, require interworking functions in the network, e.g., via a web services paradigm based on advanced service delivery platforms (SDPs) and next-generation network (NGN)
multimedia session control frameworks such as IP Multimedia Subsystem (IMS). The question now arises as to how much more bandwidth is required to support these services than was supported by the first phase “all IP” network transformation. To answer this question requires consideration of a number of factors, as follows: • The evolution of MPEG-2 to H.264-based video compression will reduce the bandwidth required per video stream by a factor of two or three. • Full support for PVR functionality in the home will increase the number of video streams per home by a factor of two or three. • The increase in user-generated content, as well as home monitoring or e-health applications, will increase the need for at least one guaranteed SDTV quality stream from each end user, evolving to multiple streams or, in the future, to a single HD-quality stream. • Peer-to-peer (P2P) applications will be employed as a complement to conventional media distribution architectures and will increase the upstream bandwidth requirement from the home, albeit the bandwidth requirement is less than for live streaming media content.
DOI: 10.1002/bltj
Bell Labs Technical Journal
5
•
Blending of “multimedia” applications will lead to an increase in the video content of all communications sessions and add multimedia interactivity to nontraditional communications endpoints and activities such as TV watching. In combination, this will result in the need for higher bandwidth and/or dynamic, differentiated QoS guarantees. • New forms of conventional/studio generated video such as 3D or immersive video will become available and will gradually evolve to become “must have” services, much as DVD-quality and HDTV have previously. To underscore the importance of some of the new media trends, an internal Alcatel-Lucent study of Internet traffic in completed in early 2007 found that in North America, HTTP traffic comprised 38 percent of network bandwidth, with YouTube (Internet video) already accounting for 4 percent of HTTP traffic. In Germany, peer-to-peer traffic accounts for 36 percent of network bandwidth while 70 percent of P2P traffic is video. The net conclusion from the preceding must be that the bandwidth and QoS requirements will likely increase by at least a factor of 2 to 4 relative to today’s “all IP” networks, both in the downstream (to the home) and upstream (from the home) directions. To accommodate this, changes in the network infrastructure are required, such as evolving to fibercentric architectures in the access network (with, for example, 80 Mbps downstream and 30 Mbps upstream on a Gigabit passive optical network [GPON]) and by the addition of more switching and routing capacity in the aggregation and core networks (up to 500 Gbps throughput, for example), as depicted in Figure 2. It is important to recognize, however, that ultimately the core network bandwidth evolution can, and will, depend on the decision by the operator regarding the placement of the application servers and multimedia content relative to the requesting users, as discussed in the following section. Intelligence at the Edge Given the massive capital expenditure by operators to move to an “all IP” network, the key, from an operator standpoint, must be to support the required
6
Bell Labs Technical Journal
DOI: 10.1002/bltj
services evolution and still leverage their new IP network, with a minimum of network upgrades. It is this requirement that points to the next phase of network evolution: the movement of content and services intelligence to the edges of the network (to prevent large-scale growth in the core network bandwidth), as well as the optimization and evolution of the access network distribution network technologies (both copper and fiber), to support the higher bandwidth needs. These trends are highlighted in Figure 4, and the six large-scale trends identified are described below: IP routing in the access node. Adding such layer 3 forwarding capabilities to a node leads to improved scalability and service security. Scalability is enhanced because fewer medium access control (MAC) addresses are exposed to the aggregation network and the limited number (4095) of virtual local area network (VLAN) labels can be reused. Service security is enhanced because the IP address is known a priori by the operator, unlike the MAC address. In addition, the ability to support complete services separation using multiple routing instances, each supporting a different set of routing protocols, allows a scalable, secure wholesale services solution to be provided. As a general matter, wholesale services support on the access node is an increasingly important requirement for many operators and may drive the design of many new features to allow multiple services instantiations or partitions in the AN. L4 subscriber/services management and resource admission control in the AN. There is an increasing need to be able to discriminate and police different services in the access node, in order to support the requisite per service, per subscriber QoS. There are two clear cases where providing such functionality higher in the network is not adequate, 1) for multimedia flows originating in the home and 2) for downstream flows from multiple wholesale service providers directly connected to the AN. In the former case, the upstream bandwidth from the residential gateway (RG) to the AN, and from the AN to the aggregation network, may be sufficiently constrained (with many services competing for the same bandwidth) that congestion will occur before the traffic hits the aggregation network. This in turn results in a
NG home network RG/ONT as home media control hub with embedded femtocell
L4 subscriber/service management and resource admission control moving to AN and RG for upstream flow policing policy enforcement Broadband service routing
IP/MPLS Optical core Broadband service access node
Optical transport Broadband service aggregation BTV VoD video servers
NG access technologies Copper network: BW target: 50 –100 Mbps DSM L1–3: gains of 10 –100% in BW improved stability Fiber network: BW target: 300 Mbps/sub NG PON: 10G TDM, 4 2.5G GPON, DWDM
AN—Access node BGF—Bearer gateway function BTV—Broadband television BW—Bandwidth DSM—Dynamic spectrum management DWDM—Dense WDM ESA—Emergency services access IMS—IP Multimedia Subsystem IP—Internet Protocol MPLS—Multiprotocol label switching
Layer 3 (routing) moving to AN for improved scalability, security and service wholesaling
Layer 4–7 video streaming, processing and storage moving to AN and aggregation network for improved BW scalability and service responsivity IMS in access : BGF for NAT traversal, VoIP service security and policing; ESA support
NAT—Network address translation NG—Next-generation ONT—Optical network termination PON—Passive optical network RG—Residential gateway TDM—Time division multiplexing VoD—Video on demand VoIP—Voice over IP WDM—Wavelength division multiplexing
Figure 4. Elements of the second phase of network transformation.
violation of the QoS requirements, so upstream flow policing has to be performed in the RG or AN to mitigate. In the latter case, the point of ingress of the traffic into the operator network is the AN, so ingress policing for downstream flows has to be provided in the AN in order for service level agreements to be enforced on a per-service provider, per-service, and per-subscriber basis. A related functionality is that of resource admission control (RAC) in the access node. A local RAC function is increasingly necessary as the number of video services (multicast, PVR, VoD, timeshift TV) and
video-capable endpoints (PCs, STBs, DVRs) in the home increases, and the AN network bandwidth remains constrained. For example, considering the per subscriber copper “loop,” even in fiber to the node (FTTN) deployments, the bandwidth supported per user might only be in the range of 25 Mbps, of which perhaps 20 Mbps is reserved for video services. With HD streams requiring on the order of 8 Mbps to 9Mbps per stream, and 1 Mbps to 2 Mbps for SDTV, it is easy to see that if two HD streams and one SD stream are being viewed, an additional stream cannot be supported and the request should be denied
DOI: 10.1002/bltj
Bell Labs Technical Journal
7
in order to prevent corruption of the existing streams. For local multicast services, the AN can be provisioned with policies that outline such constraints and can autonomously perform the RAC function. However, when one considers VoD requests, the AN is not typically involved in the control plane signaling (i.e., the stream request), so potential conflicts can still arise with the VoD server admitting the flow and the AN being unable to support the request due to the current multicast load, either per AN, or per subscriber loop. To address this situation effectively and allow for maximum revenue potential, coordination between a central RAC (that serves to aggregate the requests from applications such as VoD servers) and the local RAC in the AN (which manages the local resources) is required. This same control mechanism can also be employed to dynamically push policies to the AN and to support “bandwidth on demand,” for example. IMS in access. In addition to the RAC functionality described above, which some people consider part of IMS since they both appear in the Telecommunication and Internet converged Services and Protocols for Advanced Networks (TISPAN) NGN reference architecture [1], there is an additional set of functions that can be provided at the point of traffic ingress associated with Voice over IP (VoIP) services security, network address translation (NAT) traversal, and local call routing. Simply put, there is building momentum to take the functionality of a session border controller (SBC) and to disaggregate it into a control plane piece that is separate from the media path processing functions, in order to improve the scalability. In such a case, the media path processing (so-called gate open/close, NAT traversal, flow metering, support for lawful intercept, and IP topology hiding) should logically be provided by the AN for optimum services and network security. Furthermore, these capabilities naturally allow for the local routing of calls in case the connection to the primary Session Initiation Protocol (SIP) server is lost. Layer 4-7 video streaming, processing, and storage. It is widely accepted that video-centric multimedia services will be the dominant services in terms of network bandwidth and QoS schema. Moreover, it is anticipated that these services will become increasingly personalized, moving from a multicast to a unicast 8
Bell Labs Technical Journal
DOI: 10.1002/bltj
transport paradigm. Notably, in addition to conventional VoD, other emerging services also require unicast delivery. These services include: • Network PVR (nPVR), or user-requested storage of content in network servers, • Timeshift TV, or operator-controlled storage of a one- to two-week period of all broadcast content, for later viewing by users, • Pause live TV, or storage of a rolling window of one hour of broadcast content, allowing the user to rewind to the beginning of the current program, and • Personalized advertising, or ads that are dynamically selected and inserted on a per-user basis. Similarly, services such as fast channel change and reliable delivery typically require periods of unicast “bursting.” As such, the video load on the network will no longer be characterized by 500 to 1,000 multicast channels independent of the number of users attached to the network, but rather three to four streams per user multiplied by the number of active users, or millions of “channels.” If conventional centralized video delivery architectures are employed for all such services, a large-scale rebuild of the end-toend IP network would be required in most cases, something that network operators would clearly like to avoid. The solution is to push the video services logic to the edge of the network, so that the translation from multicast to unicast can be provided only where necessary—close to the requesting user—and not consume core network bandwidth. Similarly, for native unicast services such as nPVR and VoD, if a storage device or “cache” is provided close to the end user and this cache is used to store popular titles or programs, a significant reduction in core network bandwidth can be expected. Next-generation access technologies. With all the new video-rich services described in the preceding, it should be clear that the demands on the access network bandwidth, and in particular the bandwidth in the first mile (from the AN to the end user), will continue to increase. Yet, as stated above, the likely requirement for those operators who have already committed to a large-scale FTTN or fiber to the home (FTTH) build, is that the access network infrastructure—e.g., the location and type of outside plant cabinets and
splitters as well as the copper and fiber layout—should be conserved. New access network technologies are currently in the process of standardization and development to allow exactly this to occur; on the copper side an evolution of very high speed digital subscriber line 2 (VDSL2) technology is being investigated that would allow cancellation of all the dominant crosstalkers into a given line, resulting in potential increases in loop bandwidth of 25 percent to 100 percent depending on the specific crosstalk scenario and loop lengths. On the fiber side, an evolution of the passive optical network (PON) technology is also envisaged, in order to support four times the bandwidth per end user, at the same fiber split ratio. Both evolutions will allow the existing access infrastructure to be reused to a great extent, possibly even including the existing customer premises equipment (CPE). When these technologies are combined with the reliable delivery methods for video services, an unparalleled, high bandwidth and high quality personalized video experience should be provided to the end users. Next-generation home network. If the connection between the AN and the CPE is considered the “first mile” of the operator network, then the connections within the home are increasingly being viewed as the “first few feet” of the same network. Anecdotally it has been reported by an operator that 50 percent of the packet loss experienced by end users arises from within the home; another operator reports on the order of 8 Mbps of throughput loss within the home network (compared to that at the CPE). This situation has led to the operator deploying home network distribution technologies that are relatively noise immune, or even deploying new in-home wiring. At the same time, the typical home network is becoming more complex as it comprises an array of PCs, STBs, storage devices (PVR, media server), home control elements, gaming devices, an L3 switch-router or two, and maybe a wireless femtocell, each of which may be trying to access or enable access to services from another device either within the home or from the larger network. Not surprisingly, the complexity of provisioning and maintaining such home networks is beyond the scope of most users, so there is a new trend for the operator not only to provide
installation services, but also to support automated provisioning and troubleshooting/diagnostics of the home network.
Summary In summary, this issue of the Bell Labs Technical Journal is dedicated to a discussion of all the above topics of relevance to next-generation access. It is a very timely issue as the publication year—2008— should see many of these trends becoming a reality, at least at the level of trials in operator networks. So, the “proof of this pudding” should not be long in the tasting! *Trademarks YouTube is a trademark of Google, Inc.
Reference [1] European Telecommunications Standards Institute, “Telecommunications and Internet Converged Services and Protocols for Advanced Networking (TISPAN), Common Basic Communication Procedures, Protocol Specification,” ETSI TS 183 028, v1.1.1, Apr. 2006, . (Manuscript approved January 2008) MARCUS K. WELDON is the chief technical officer (CTO) of Alcatel-Lucent’s Fixed Access Division (FAD) and is based in Murray Hill, New Jersey. In this role, he and his CTO team are responsible for defining the strategic portfolio directions to be followed in the access space (including home networks), as well as for Alcatel-Lucent’s IPTV middleware and video content distribution solutions. The FAD CTO team synthesizes this technology strategy by leveraging their vast technical capabilities, and through extensive interaction with the Bell Labs Research community, the Network Technology and Strategy group in corporate CTO, and through many discussions with network operators. Prior to becoming the CTO for FAD in Alcatel-Lucent, Mr. Weldon was the CTO for Broadband Solutions in Lucent Technologies, which also comprised the access network portfolio and IPTV middleware solutions. He assumed this position while still a technical manager and member of technical staff in Bell Laboratories. He received a B.S. in chemistry from Kings College, London, United Kingdom, and a Ph.D. in physical chemistry from Harvard University, Cambridge, Massachusetts.
DOI: 10.1002/bltj
Bell Labs Technical Journal
9
THIERRY VAN LANDEGEM leads the Alcatel-Lucent Bell Labs Research and Innovation team. He is based in Antwerp, Belgium. Mr. van Landegem joined Alcatel-Lucent as a research engineer working in various areas including broadband ATM, network survivability, data communications, and mobile communications. He has been active in several international research projects and standards organizations in the fields of ATM and broadband ISDN. During his tenure, he has been responsible for network strategy defining network architectures for the Alcatel group and has served as Internet Access division vice president, with business and strategy responsibilities in the broadband access domain. Mr. van Landegem later joined the Chief Technology Office and Research and Innovation Department as location director for Antwerp, Belgium, and was responsible for research and innovation activities, network strategy, and patents. He was awarded a master’s degree in sciences (electrical engineering) from the University of Brussels and a master’s degree in business administration from the University of Leuven, both in Belgium. He has written over 50 papers and twice has served as guest editor for IEEE journals. EDWARD S. SZURKOWSKI is vice president of the Wireless and Broadband Access Research Center at Bell Laboratories in Murray Hill, New Jersey. He is responsible for a broad range of research programs in wireline access, wireless access, digital video systems, and the integration of these technologies. Previously he led research efforts in metro optical systems, interactive television, electronic publishing, highperformance processors, and other networking and computing systems. Mr. Szurkowski received a B.E.E., M.E.E., and Ph.D. in electrical engineering from the University of Delaware, Newark. ◆
10
Bell Labs Technical Journal
DOI: 10.1002/bltj
