`
`Why ReFLEX has become the industry standard for wireless data delivery
`
`subscribers – greater than 2/3 of the existing messaging
`industry subscriber base.
`
`
`
`Broadest Nationwide Network Coverage
`
`Experience has shown that coverage is perhaps the
`most
`important
`requirement
`for
`broad-based
`deployment of wireless data applications to both
`businesses and consumers. Wireless service must cover
`users where they live, work, travel and vacation.
`ReFLEX networks already in use cover more square
`miles, including suburbs and recreation areas, than any
`other terrestrial wireless data network of any kind in
`the United States. ReFLEX coverage goes beyond
`major metropolitan areas and business centers to
`include hundreds of small cities and fast growing
`suburbs.
`
`ReFLEX network coverage does not end at the U.S.
`borders. Canada and Mexico, the two most traveled
`countries by Americans, are also building out ReFLEX
`networks. Bell Mobility, a wholly owned subsidiary of
`Bell Canada, is building the Canadian ReFLEX wireless
`data coverage which will be commercially available
`before the end of the year. Telefonos de Mexico
`(Telmex), the largest telecommunications company in
`Mexico, is building in Mexico and is expected to have
`full coverage by mid 2001.
`
`
`
`Better Subscriber Equipment – Lower Cost and
`Broader Selection
`
`ReFLEX protocol is optimized for high-speed wide
`area wireless data delivery using small, inexpensive,
`always-on,
`battery-powered
`subscriber
`units.
`Manufacturers are developing and shipping a greater
`variety of subscriber equipment for ReFLEX networks
`than for any other kind of wireless data network. There
`are multiple manufacturers of equipment for this
`market, and more are expected to enter the ReFLEX
`market soon. This diversity provides a greater range of
`existing messaging units from which customers can
`
`
`
`1
`
`TABLE OF CONTENTS
`
`Executive Overview............................................1
`ReFLEX Network Design .................................2
`ReFLEX Network Coverage...........................15
`ReFLEX Network Capacity.............................18
`Summary/Conclusion.......................................20
`
`
`
`
`
`EXECUTIVE OVERVIEW
`
`ReFLEX protocol based packet data networks are
`becoming the industry standard for wireless data
`applications because of their economic performance
`advantages and the fact the protocol is designed for
`long term migration to higher speeds and capacity as
`demand requires.
`
`
`
`Industry Standard Wireless Data Protocol
`
`Industry analysts are projecting an explosive growth in
`wireless data users over the next few years driven
`primarily by consumers’ needs for access to Internet
`based data (e-mail, instant messages and information
`on demand) and wireless e-commerce. ReFLEX
`networks are a preferred platform for these wireless
`data applications because, as described in this paper,
`they are engineered to deliver low cost, low bit error
`rate (BER) data over vast geographic footprints using
`small,
`inexpensive, always-on mobile devices that
`operate for weeks on a single AA battery.
`
`In addition, ReFLEX networks are generally operated
`by companies focused on the single purpose of reliably
`developing, deploying and supporting wireless data
`applications. USA Mobility, MCI/SkyTel, PageNet,
`Arch Communications, USA Mobility, AirTouch and
`TSR Wireless all have investments based on the
`Motorola-developed ReFLEX
`protocol. These
`companies represent more than 30 million U.S.
`
`General Electric Co. 1017 - Page 1
`
`
`
`major national retailer like Radio Shack is a good proxy
`for the ability of a wireless data network to be mass
`marketed to consumers nationally.
`
`
`
`REFLEX NETWORK DESIGN
`
`
`
`ReFLEX Network Protocol
`
`The ReFLEX protocol is uniquely engineered for
`reliable delivery of wireless packet data. ReFLEX has
`been selected by more carriers as the industry standard
`platform for light and medium wireless data load
`applications, such as e-mail, instant messaging, e-
`commerce, GPS locating, and Internet information on
`demand. Every commercially operating narrowband
`PCS
`licensee has committed
`to supporting
`the
`common version 2.7 standard ReFLEX wireless data
`protocol. In the authors’ opinion, all will likely migrate
`to the new third generation (3G) ReFLEX as required
`over the next several years.
`
`ReFLEX networks use state-of-the-art, software-
`configurable base stations that support multi-mode,
`multi-frequency communications. This allows carriers
`to migrate network
`assets,
`infrastructure
`and
`consumers
`to higher
`speeds
`and
`subsequent
`generations of ReFLEX as capacity needs and
`applications evolve. ReFLEX networks support data
`features such as token and broadcast messaging that
`hybrid voice and data networks will not or cannot
`implement in the foreseeable future.
`
`One key benefit of ReFLEX network technology is its
`implementation of wireless “mesh networking” at the
`base station level that permits mobile devices to
`communicate bi-directionally with multiple base
`stations simultaneously. This delivers unprecedented
`always-connected service and coverage reliability for
`mobile devices compared to the single base station,
`single link implementation of all other packet data
`networks such as broadband PCS, Mobitex and RD-
`LAP networks.
`
`
`
`
`2
`
`
`
`choose and ensures access to subsequent generations of
`low-cost, consumer as well as business-oriented
`wireless data devices.
`
`The newest device just coming to market is Motorola’s
`Talkabout T900, expected to be broadly available in
`August of this year. The T900 is the smallest, lowest-
`cost and most consumer-friendly wireless e-mail and
`wireless instant messaging device on the market. The
`T900 will be a catalyst for consumer acceptance of “2-
`Way I-Messaging” services.
`
`Glenayre provides two other subscriber device options.
`The Accesslink II
`is the
`lightest weight 2-way
`messaging device on the market. Messages are created
`on the easy to use virtual keyboard, allowing for one-
`handed operation and quick navigation through the
`device's
`intuitive folder-based user
`interface. The
`AccessLink II serves double duty as a wireless
`connection for PDAs via the infrared ports. The
`@ctiveLink is a 2-way wireless messaging plug-in
`module that enables the Handspring Visor to become a
`mobile Personal Information Management system.
`
`No other wireless technology has comparable device
`performance and price points nor is expected to have
`in the foreseeable future. ReFLEX wireless data
`technology is also expected to become embedded by
`several manufacturers in personal digital assistants
`(PDAs) and in machine data modules for telemetry
`applications.
`
`
`
`Massive Retail Distribution to Consumers
`
`ReFLEX networks provide wireless data coverage to
`more retailers, than any other wireless data network in
`the U.S. For example, ReFLEX networks cover 90
`percent of all RadioShack’s 5,000 corporate-owned
`stores, creating an unparalleled retail environment
`capable of delivering the lowest-cost consumer device
`for wireless e-mail and wireless instant messaging.
`Comparatively, the Sprint PCS network, the largest
`single broadband footprint in the country, covers only
`2/3 of these stores while the BellSouth Mobitex
`wireless data network covers even less. Coverage of a
`
`General Electric Co. 1017 - Page 2
`
`
`
`The designers of ReFLEX have adopted a far simpler
`address model, which does not embody any implicit or
`explicit notion of network or sub-network. This implies
`the complete ability of the mobile device to roam
`between serving areas of one or more service providers
`without modification to the mobile unit’s address. That
`is, the ReFLEX address of a mobile unit is a global and
`intrinsic attribute, in contrast to the IP address of a
`networked host, which must be adapted to its current
`network location.
`
`IP is a “balanced” peer-to-peer protocol. In contrast,
`ReFLEX supports an “unbalanced,” host-to-terminal
`model. Said another way, ReFLEX devices do not
`have the inherent capacity to communicate with other
`peer units. Rather, they must always inter-work through
`the infrastructure of the messaging network, which
`provides store-and-forward message delivery functions
`that place a substantial onus on the sending host in a
`“balanced” protocol.
`
` A
`
` very important mobile data requirement is that any
`battery-powered mobile unit must not unnecessarily
`bear the burden of message delivery, as it does under
`the Transport Control Protocol (TCP). In an IP
`network, using TCP as the method for assuring
`message delivery between
`two mobile units,
`intermediate routing systems provide no inherent store-
`and-forward functionality. If the sender is in good
`coverage and the receiver is in poor coverage, TCP
`would place the onus for retries on the sending mobile
`device. The common physical link and network layer
`protocols – Ethernet, IEEE 802, and IP – provide no
`mechanisms to compensate for the vagaries of UHF
`mobile radio interface.
`
`The introduction of store-and-forward agents in the
`network creates a fundamental
`imbalance
`in the
`communication pathway between communicating end
`units. Almost everyone will be familiar with this effect
`in the context of simple mail transfer protocol (SMTP)
`and post office protocol (POP) servers. While it would
`be perfectly feasible for two Internet hosts to send
`messages directly using SMTP, this is almost never
`done. Instead, mail servers are established and the end
`hosts communicate with the mail servers using a POP.
`
`
`
`3
`
`
`
`The table below shows the plan for ReFLEX protocol
`migration subject to adjustment by the carriers,
`Motorola and Glenayre as required based on actual
`network loading and user requirements.
`
`
`
`ReFLEX *
`Version
`
`Mesh
`Simulcast
`
`Channels/
`50Khz
`
`Kbps/50k
`Hz
`
`ReFLEX 2.0
`4FSK
`
`ReFLEX 3.0
`4QAM
`
`ReFLEX 3.0
`16QAM
`ReFLEX 3.0
`256QAM
`
`Yes
`
`Yes
`
`Yes
`
`No
`
`4
`
`14
`
`14
`
`14
`
`25.6
`
`78.4
`
`156.8
`
`313.6
`
`
`
`
`
`Cost/User
`
`FLEX 1.6, 1.8
` FM 1-Way
` Simulcast
` 6.4 kbps
` 25 KHz
` Numeric
` Limited Alpha
`
`
`
`
`
`ReFLEX Transport Layering Steps
`Source: Motorola “Get Connected”, 1998
`
`FLEX 1.9
` FM 1-Way
` Simulcast
` 6.4 kbps
` 25 kHz
` Numeric
` Limited Alpha
`
`ReFLEX 25
`FM 2-Way
` Simulcast &
`Targeted
` 19.2 kbps
` 50 KHz
` Alpha
` Limited Voice
`
`ReFLEX 3.0
`4 to 16 QAM 1 & 2-Way
` Simulcast & Targeted
` 25 kHz: 28 to 56 kbps
` 50 kHz: 78.4 to 156.8 kbps
` Voice, Data & Graphics
`
`ReFLEX 3.0 - Warp
` 4 to 256 QAM 2-Way
` Targeted
` 25 kHz: 28 to 112 kbps
` 50 kHz: 78.4 to 313.6 kbps
` Voice, Data & Graphics
`
`
`The ReFLEX Protocol Design
`
`Any modern communications network must be
`designed cognizant of the Internet and its protocols.
`This is certainly true of any wireless data network, since
`most of the “off-net” traffic will flow either from the
`Internet or from a private network using Internet
`protocols. The architects of wireless networks
`invariably are forced to come to grips with two aspects
`of the Internet Protocol (IP). First, its address plan is
`verbose and does not accommodate mobility. Second,
`its protocols are verbose and depend upon its address
`plan.
`
`General Electric Co. 1017 - Page 3
`
`
`
`Three broad models of interconnect to the Internet are
`available: a bearer service, a
`teleservice, and a
`teleservice supplementary. In the bearer service model,
`it is assumed that the Internet host has the capability to
`encode and decode FLEXsuite. In this case, pure
`binary content is provided to the ReFLEX network,
`which transports the content untouched to the mobile
`device2. In the teleservice case, an Internet protocol,
`say SMTP, is mapped onto a FLEXsuite protocol, for
`example the FLEXsuite protocol mailto. In the
`teleservice supplementary case, an Internet protocol,
`again say SMTP,
`is mapped onto a FLEXsuite
`protocol, for example WEM, and in addition, the
`ReFLEX network provides supplementary services
`such
`as mail-box
`filtering,
`store-and-forward
`guarantees, message compression, mail re-routing, and
`attachment stripping.
`
`
`
`ReFLEX Air Interface
`
`At the air interface, ReFLEX supports guaranteed
`communications at the physical, link, and network
`layers. In terms of the data link layer, ReFLEX uses a
`pair of simple error detection and correction methods,
`far simpler3 than those employed
`in cellular or
`wideband PCS protocols. In contrast to cellular air
`interfaces, ReFLEX employs a much more complex
`simulcast channel4, on the forward (network-to-mobile
`device) path, and an adaptive diversity channel on the
`reverse (mobile device-to-network) path. Likewise, the
`forward channel power levels used in ReFLEX are
`typically an order of magnitude stronger than in cellular
`systems. The conclusion is that ReFLEX radio links
`usually operate at a lower raw BER than cellular links in
`the same location. Also ReFLEX forward channels will
`
`
`2 Currently, the ReFLEX industry is moving to adopt common
`protocol – wireless communications transfer protocol (WCTP) –
`for the delivery of binary content.
`3 The forward link protects user data with a (21,32) BCH code. The
`reverse link protects data with Reed Solomon code (31,23).
`Cellular systems typically use layered error protection codes with a
`rate 1:2 or rate 1:4 convolutional code, at a minimum, on top of
`some further cyclic or block code for error detection.
`4 A typical ReFLEX air interface involves dozens of transmitters,
`timed by GPS, and optimized with offsets obtained by simulated
`annealing.
`
`
`
`4
`
`
`
`This has the distinct advantage of allowing mail to be
`relayed through various post offices to the one closest
`to the end host without that final host even having to
`be active on the network. It has the disadvantage of
`increasing message latency by virtue of the delays
`through the relay network and in the need for the
`receiving host to poll its POP server.
`
`In a low-latency mobile messaging network, in which
`communicating end units are in good coverage and
`active on the network, it would be undesirable to force
`mobile units to poll servers for messages. This implies a
`set of network functions that has
`little,
`if any,
`comparison in the Internet.
`
`
`
`ReFLEX and the Internet
`
`It is absolutely necessary to inter-operate effectively
`with Internet protocols such as SMTP for email, http
`for web traffic, and so on. Again, the designers of
`ReFLEX have accounted
`for
`this
`requirement.
`ReFLEX supports a recursive stack model, which is in
`some respects even more sophisticated than the linear
`stack models of TCP/IP and ISO OSI. The protocols
`of the ReFLEX stack are collectively referred to as
`FLEXsuite. For most of the dominant Internet
`protocols, there exists at least one corresponding
`FLEXsuite protocol type1. FLEXsuite also supports
`the common set of MIME types as well as the wireless
`applications protocol (WAP) extensions.
`
`All of this taken together implies a capability to
`transport arbitrary binary content from an Internet host
`to an arbitrary application running on a ReFLEX
`mobile device, and to have that application recognize
`how to process the data based upon content identifiers.
`This is the essence of the present success of the
`Internet–the ability for applications on one host to
`transport arbitrary content to applications on another
`host with absolute disregard for the
`intervening
`network elements.
`
`
`
`
`1 For SMTP e-mail, there are two: mailto and wireless e-mail
`(WEM). Mailto supports a simple interface. WEM supports full
`RFC822 headers and corresponding functionality.
`
`General Electric Co. 1017 - Page 4
`
`
`
`messages to the mobile and those that are responsible
`for receiving reverse channel traffic from it. As an
`extension, there is no necessary relationship between
`receiver site
`locations, antenna patterns, antenna
`heights, and so on, and those same attributes of
`transmitter sites5. In fact, it is generally desirable to
`offer a somewhat larger coverage footprint on the
`reverse channel than on the forward channel, to
`guarantee the mobile unit’s ability to contact the system
`whenever it sees forward channel coverage.
`
`In short, the RF design of the forward channel and
`reverse channel can be quite distinct in ReFLEX. Based
`upon design choices made by early implementers of the
`protocol, there has come to be an erroneous view that
`ReFLEX networks must have a much higher density of
`receive sites than transmit sites on the ground. This is
`not true, and recent
`implementations have been
`constructed with receive to transmit sites at a 1:1 ratio.
`
`
`
`ReFLEX Network Capacity Design
`
`ReFLEX networks are capable of supporting increased
`offered load using an approach similar to cell splitting
`in cellular systems. This involves dividing a region that
`would otherwise be a simulcast zone in a traditional
`messaging network into distinct sub-zones, each with
`its own forward and reverse channel frequency
`assignments. This allows a ReFLEX network to have
`capacity growth in both directions similar to a cellular
`system, by reducing the effective area of coverage of a
`serving region. In cellular, this is a cell. In ReFLEX,
`this is a sub-zone6. Likewise, ReFLEX sub-zones do
`not all have to have the same capacity for load
`handling. One may have only a single forward and
`reverse channel, while a neighboring sub-zone may
`have several. And variations of the ReFLEX protocol
`supported in sub-zones do not all have to be alike.
`Therefore, dense regions of offered load can be served
`with high-density, high-capacity implementations of the
`
`
`5 This allows ReFLEX zones to be configured as “two-way”,
`“partial coverage” or “one-way”; a feature that we’ll discuss later
`in this document.
`6 The number of sites in a sub-zone is highly flexible. It might
`include all sites in the parent zone or consist of only one site.
`
`
`
`5
`
`
`
`show less correlated bit error patterns than cellular in
`similar locations.
`
`So the usual cellular system operates on a Rayleigh
`BER curve versus carrier to interference. And it is in
`the interest of cellular service providers to maximize
`their network capacity by operating their voice services
`as close as possible to the poorest raw BER that they
`can afford. This maximizes the frequency re-use in
`their cellular plan. It also implies a poor starting point
`for a high-grade data service. In order to compensate
`for this, it is conventional to add high levels of error
`protection.
`
`In fact, such extreme levels of error protection are
`necessary under the most marginal of link conditions
`that there is almost no user bandwidth left. Therefore,
`these networks are usually made adaptive in the sense
`that the link quality is monitored and the channel
`coding is altered on the fly to maximize user data
`throughput. Unfortunately, this optimization is difficult
`for short data bursts. It works best for a continuous
`data stream, so that the adaptation logic has something
`to work with. This makes any interpretation of average
`or worst-case user data throughput characteristics
`challenging at best.
`
`This trade-off of user bandwidth for coding protection
`in cellular systems is interesting in contrast to the
`design philosophy of ReFLEX. Since ReFLEX does
`not start with a connection-oriented voice component,
`it could be designed for high performance, bursty,
`short data messaging from the start. The contrasting
`assumption in ReFLEX is that the mobile device is
`either available to the system, or not. If not, as proven
`by the failure to deliver a message, then the system
`begins to search for the device. Full details of the
`search process are beyond the scope of this paper, but
`suffice it to say that the mobile device is either
`recovered by the network or it autonomously registers
`again. In either case, any pending messages are
`subsequently delivered at full speed.
`
`It bears mentioning that in ReFLEX there is only a
`loose relationship between the base sites that are
`responsible for the delivery of forward channel
`
`General Electric Co. 1017 - Page 5
`
`
`
`
`
`a major
`regions of
`protocol, while outlying
`metropolitan area can be served with a capacity
`commensurate with the offered load.
`
`Returning to the trade-off of user bandwidth for error
`protection
`in cellular systems, we can tie error
`protection in ReFLEX to a similar trade-off; namely,
`the number of diversity channels in a sub-zone. The
`philosophy in ReFLEX is to use diversity instead of the
`complex error coding typical of cellular. Since there is
`no capacity driver in ReFLEX so strongly associated
`with frequency re-use, there is also no need to choose
`an operating point for the network with the absolute
`maximum of survivable raw BER as in cellular. This
`implies that cellular networks must claw their way back
`to an acceptable user BER with coding, sacrificing user
`data rate in the process. ReFLEX networks are meant
`to begin at an acceptable error rate at all points within
`the serving area.
`
`Another aspect of the ReFLEX architecture is that the
`aggregate signaling rates on the forward and reverse
`channels are highly de-coupled. ReFLEX allows
`forward channel signaling rates of 1,600, 3,200, and
`6,400 bits per second. Binary frequency shift keying
`(FSK) is used for 1,600 bps and can be used for 3,200
`bps. Four-level FSK is used for 6,400 bps and can be
`used for 3,200 bps as well. On the reverse channel,
`ReFLEX allows rates of 800, 1,600, 6,400 and 9,600
`bps. Four-level FSK is used for all reverse channel
`transmissions.
`
`the aggregate
`This doesn’t completely describe
`signaling rate in a market, however. In the forward
`direction, a number of transmitter sites will be
`synchronized for simulcast operation within a sub-
`zone. So all these transmitters are occupied with the
`same forward channel information stream, perhaps at
`6,400 bps. Any given user might receive only part of
`that stream, but all devices would monitor the same
`shared
`information resource(s)7. On
`the reverse
`channel, mobile units that need to access the network
`
`7 A ReFLEX sub-zone can support several outbound channels.
`However, once a device is “camped” on a channel, it remains there
`until some event occurs that would cause it to drop off the air
`interface or chose a better channel.
`
`compete with one another on a part of the available
`bandwidth dedicated for this purpose. Since the
`transmit range of a mobile unit may be limited to only a
`few fixed sites within a zone, the aggregate raw
`information rate for all sites in the market will be the
`sum of activity at all of them. The aggregate non-
`redundant information rate will be less than this
`because of duplication in information received at
`multiple sites and representing the same mobile unit’s
`transmissions.
`
`Once a mobile station has succeeded in contacting the
`network on the contention part of the reverse channel,
`it may, for example, register with the network, or
`request specific
`information services, or request
`bandwidth for a long inbound message. If we follow
`the last case through, the network will validate the
`subscriber account, at the corresponding MS-H, and
`then send a forward channel command to the mobile
`unit that tells when to transmit the message. This
`allocation will be made in a part of the reverse channel
`that is reserved for such scheduled activity, in contrast
`to the contention access part. This raises several
`possible scenarios for the allocation of this scheduled
`bandwidth within a zone or sub-zone. In general,
`allocations within a sub-zone will potentially be
`“blocking.” That is, if two mobiles were scheduled at
`the same time, their messages could interfere with one
`another destructively at one or more receiving sites.
`
`The network may “overbook” the scheduled reverse
`channel
`resource when
`it determines
`that
`two
`subscriber units will not interfere with each other. An
`alternative capacity enhancing approach is to reduce the
`number of sites in a sub-zone to the point where
`overbooking would add little benefit. Part of the
`flexibility of ReFLEX is that it allows either, or neither,
`approach.
`
`increasing
`the above-mentioned capacity
`All of
`solutions for both the forward and reverse channels
`need to be carefully orchestrated by the ReFLEX
`infrastructure equipment. Seamless sub-zoning and
`overbooking algorithms are collectively known as
`“software-based” capacity increasing solutions. Until
`the next generation protocol
`is commercialised,
`
`
`
`6
`
`General Electric Co. 1017 - Page 6
`
`
`
`In fact, the most extreme source of message delay is
`likely to be due to the link recovery processes in the
`event of a device being unavailable for communication.
`In current ReFLEX networks, these recovery processes
`are managed by network elements in the Wireless
`Message Transport Protocol
`(WMtpTM)
`reference
`model, developed by Glenayre. The
`two most
`important of these are the Output Messaging Switch
`(MS-O) and the Home Messaging Switch (MS-H). It is
`the role of the MS-O to manage the mobility of a
`device within the scope of a serving market, often
`called a zone. It is the role of the MS-H to manage the
`mobility of a device across all of the zones of a service
`provider, including roaming onto the zones of another
`service provider. This MS-H management role includes
`both the grant and denial, for any reason, of service
`rights to a mobile device attempting to operate within
`an MS-O.
`
`To those familiar with cellular networking, the role of
`the MS-H is closest to the combination of the Mobile
`Switching Center (MSC) and its associated Home
`Location Register (HLR). Likewise, the MS-O is
`functionally closest to the Visitor Location Register
`(VLR) of cellular networks. One notable exception is
`that in the WMtp reference model, all mobile devices
`are visitors in every possible MS-O. In fact, roaming in
`a WMtp-based ReFLEX network has little, if anything,
`to do with location. Rather, it has to do with the
`matching of service provider identifiers stored in the
`mobile unit, broadcast on the air
`interface, and
`advertised by WMtp elements.
`
` A
`
` subscriber device is associated with only one MS-H,
`which is the anchor point for all forward and reverse
`channel messaging to it and from it. The MS-O acts as
`intermediary agent between the MS-H and the mobile
`unit. In the event of a communications failure at the
`level of an MS-O, the MS-H is informed of the failure.
`The original MS-O, and perhaps others, begins a search
`for the device. Recovery of the device by the original
`MS-O or by another one is signaled to the MS-H, and
`message
`delivery
`recommences. MS-O
`search
`procedures are adaptive9, and work with a variety of air
`
`9 An MS-O can be one of three classes: two-way, partial-coverage,
`or one-way. In a two-way zone, forward and reverse coverage is
`
`
`
`7
`
`
`
`software-based capacity increasing solutions represent
`the most cost-effective method of increasing the
`carriers’ return on the investment on their frequency
`spectrum and network equipment. It is estimated that,
`when all of the ReFLEX 2.7-based software capacity
`enhancement is implemented, a nation-wide ReFLEX
`carrier will have the option of increasing the network
`capacity by at least 10 times over the ReFLEX
`networks currently deployed.
`
`
`
`ReFLEX Store and Forward Latency
`
` A
`
` comparison and contrast of messaging latency in
`ReFLEXTM and cellular networks is a daunting task.
`Both types of wireless network face similar issues in
`inter-operation with the dominant Internet messaging
`protocol, SMTP. The most significant issue here with
`regard to latency is the aggregate of delays through the
`various forwarding agents and firewalls that are
`involved in Internet e-mail. The response of the
`ReFLEX
`industry
`to
`this
`issue has been
`the
`development of WCTP, a session-oriented protocol for
`messaging that employs an XML-tag protocol over
`http-push. Most
`Internet-sourced messages with
`mobile devices as destinations will likely employ SMTP
`up to a gateway maintained by the wireless service
`provider. Once traffic is received by the service
`provider’s network, the inherent latency to forward the
`message to the appropriate radio link location will
`undoubtedly be slight in any well-maintained network8.
`
` A
`
` more interesting question is how will the network
`decide what the most appropriate location will be?
`With what accuracy? And with what handling in the
`event of a delivery failure? None of these issues are
`resolvable in the air interface protocol itself. Rather,
`they must be addressed in the design of network
`elements, and in the protocols for their inter-operation.
`Although somewhat obvious, a message cannot be
`delivered to a mobile device that is off or out of the
`serving area or
`incapable of
`successful
`radio
`communication with the network for any other reason.
`
`8 Although it bears mentioning that the “control collapse” of the
`ReFLEX network will also have a bearing on the message latency
`both inbound and outbound.
`
`General Electric Co. 1017 - Page 7
`
`
`
`
`
`interface parameters that are designed to ensure
`continuous and robust link availability.
`
`Cellular and wideband PCS networks do not support
`much in the way of search. Rather, the onus is placed
`upon the mobile device to monitor and update its
`location relative
`to
`the fixed network. This
`is
`accomplished through a variety of time-based or
`location-based registration procedures. A specific
`example may be instructive. Consider a mobile device
`that happens to be momentarily out of coverage
`because, say, the user is in the sixth sub-floor of an
`underground parking garage. At this point, the network
`forwards a message, and in spite of any re-tries, it
`receives no acknowledgement from the mobile. In a
`voice-based cellular network, this would represent a
`missed call; and that call attempt is lost. Any search for
`the mobile is pointless. From the point of view of the
`cellular mobile, it may have lost contact with the
`network for a period, but it has no recognition of any
`failed message event. It is within the same serving
`region when it regains contact as it was when it lost
`contact. It will not register with the network on that
`count. The clock will tick inside the mobile until its
`registration timer elapses, and then it will register again.
`At this point, any pending traffic will be forwarded to
`it. Since the registration timers impact the behavior of
`all mobiles within the network, a reduction in the time
`to deliver any pending traffic to the mobile can only be
`achieved at the cost of increasing the polling rate of
`time-based registrations from all mobiles.
`
`In contrast, in a ReFLEX network, there is a system
`parameter called the “incommunicado delay time.” If a
`mobile unit loses contact with the network for longer
`than this time, it is forced to register. In the messaging
`scenario just described, the mobile is either out of
`contact for a period less than or greater than the
`incommunicado delay. If it is less than this time, the
`system begins the search process, recovers the device
`when it is back in coverage; and the pending message is
`
`
`guaranteed to be equal. In a partial coverage zone, forward
`coverage may exceed reverse coverage. A one-way zone requires
`no reverse channel at all. Initial message delivery, any subsequent
`message re-try, and device search methods are all highly dependent
`upon this aspect of MS-O configuration.
`
`immediately sent. If the loss of contact is greater than
`the incommunicado delay, then the mobile is forced to
`register, contact is established; and again, the pending
`message is immediately sent. Only devices that have
`been proven to be unavailable to the network by virtue
`of a failed messaging event or that have detected lost
`contact are subject to this search and registration
`process10. In summary, ReFLEX makes the process of
`short message delivery event-driven instead of polled.
`
`These observations are also true of many packet data
`networks