mcgraw hill wireless data demystified phần 8 potx

Since both client wireless data bytes and datasource-to-sink control information are encoded into the 8B/10B codes,efficient transport of these protocols through a public transport netwo

Chapter 16: Packet-over-SONET/SDH Specification

sense that the latter is focused on wireless data transport only net directly over fiber is currently the interface of choice) for metro-areaapplications In contrast, besides wireless data applications, the hybridservices support TDM applications as well, and can be used in both MANand WAN scenarios

used to distinguish 256 data streams within a single SONET/SDH path

OC-1/3/12

FICON/

ESCON GBE

Path switching (PSW) Virtual concatenation

PKTSW

OC-48 LCAS

GFP TGFP

GFP: Generic Framing Procedure TGFP: Transparent GFP LCAS: Link Capacity Adjustment Scheme (b)

Next-generation SONET/SDH

DOS node

DOS node

DOS node

DOS node

DOS node

Private network

Private network

Shared bandwidth (e.g., STS-3c-10v)

Figure 16-5b shows a typical functional architecture of a DoS node.

The node provides transport interfaces, such as legacy SONET/SDH,ESCON/FICON, and GbE, for a wide range of applications The DoSnode uses virtual concatenation and GFP as enablers to efficiently packwireless application data into SONET/SDH frames It also uses LCAS toregulate the amount of bandwidth assigned to transport the client wire-less data

DoS nodes are designed to provide a wide variety of line interfaces sothat new services can be launched without deployment of new nodes

STM switch

Transport node Tributary

STM IF

STM IF STM

IF Packet IF

Packet IF

Packet switch

OC-3

OC-3

OC-12 DS3

Ether MAC IP

PPP IP

PPP IP

L2 detect destination search Schedulingshaping

Chapter 16: Packet-over-SONET/SDH Specification

New line interface cards are installed as need arises Interfaces for adata center (ESCON, FICON, Fibre Channel) and digital video (DVB-

ASI) are also utilized Figure 16-5c illustrates the hardware

architec-ture of a DoS node with Layer 1⁄2hybrid switch capability The node iscomposed of the following modules

Switch modules:

STM switchPacket switchAggregate interface cards:

OC-48/STM-16OC-192/STM-64OC-768/STM-256Tributary interface cards:

Ethernet (10M/100M/1G)Fibre Channel, ESCON/FICONDVB-ASI (video interface)POS (OC-3/STM-1, OC-12/STM-4, OC-48/STM-16)ATM (OC-3/STM-1, OC-12/STM-4, OC-48/STM-16)TDM (OC-3/STM-1, OC-12/STM-4, OC-48/STM-16, DS1, DS3,etc.)3

Node-to-node trunks are terminated on an aggregate interface card Onthe receiver side of the aggregate interface, TDM traffic continues to beswitched to either the tributary interface cards or aggregate interfacecards, while the data traffic on virtually concatenated channels is routed

to the packet switch The packet switch performs termination of virtuallyconcatenated payloads to produce GFP streams at the switch input ports

At the output ports of the packet switch, the virtual concatenation tion maps the GFP streams into virtually concatenated payloads, whichare sent to the STM switch The packet switch performs the switching ofGFP frames between ports, some connected to tributary interface cardsand the rest to aggregate interface cards, through the STM switch At thetributary interface card, GFP frames are terminated to extract the origi-nal data stream, which is then mapped to the appropriate Layer 1 and 2protocols

func-In the wireless data transmission direction, the incoming Layer 1 and

2 protocols are terminated, and wireless data streams are encapsulatedinto GFP frames at the tributary interface cards If the line interfacecard happens to have several ports, the GFP frames from the variousports are aggregated and sent to the packet switch The packet switchthen switches GFP frames, maps the frames into virtually concatenatedpayloads, and sends them to the aggregate interface cards

393

GFP Point-to-Point Frame Application

The structure of a GFP linear (point-to-point) frame is depicted in Fig 16-6.3

A typical application of a GFP linear frame is point-to-point connectionand concentration For example, data streams from multiple tributaryinterface cards can be aggregated into a same aggregate interface card.The 8-bit channel identifier (CID) in the GFP extension header is used toindicate one of 256 data streams If the available bandwidth of the aggre-gate interface is below the sum of peak traffic of all data streams, statisti-cal multiplexing is introduced to achieve concentration

The optional payload FCS field in the GFP frame can be used for formance monitoring of an end-to-end GFP path The area covered byFCS is the payload information field only, which contains the wirelessuser data Therefore, at intermediate nodes, recalculation of FCS is notnecessary, so that FCS is retained throughout the path The end-to-endpath monitoring can be used for path quality management as well as fortriggering protection mechanisms

per-SAN Interconnection by Transparent GFP

SAN deployment for disaster recovery applications has recently received alot of attention This application requires direct connection of SAN inter-faces to a WAN in an efficient manner

The conventional method for supporting this application is to simplyassign one wavelength to each SAN interface This method is inefficient interms of bandwidth usage because the SAN bit rate is generally much lessthan the wavelength modulation rate Better efficiency is achieved by mul-tiplexing several SAN signals into a SONET/SDH-modulated wavelength.Transparent GFP (TGFP) allows transparent transport and multiplexing

of 8B/10B clients such as Fibre Channel, ESCON, FICON, and DVB-ASI(digital video), as mentioned earlier Transparency means that wirelessdata and clock rate received at the TGFP ingress node can be recovered atthe egress node over a SONET/SDH network TGFP can be seen as a kind

of sublambda technique for 8B/10B interfaces over SONET/SDH (see Fig 16-7).3

An additional benefit of this solution is that TGFP provides 6.25 to16.25 percent bandwidth reduction from the original 10B rate Table 16-1shows typical VC path capacity required for SAN client transparenttransmission.3

Finally, let’s look at why generic framing procedure (GFP) is a newstandard that has been developed to overcome wireless data transportinefficiencies or deficiencies with the existing ATM and packet overSONET/SDH protocols Transparent GFP is an extension to GFP devel-

Chapter 16: Packet-over-SONET/SDH Specification

oped to provide efficient low-latency support for high-speed WAN cations including storage-area networks Rather than handle wirelessdata on a frame-by-frame (packet-by-packet) basis, TGFP handles block-coded (8B/10B) character streams The next part of the chapter describesthe GFP protocol along with technical considerations and applications fortransparent GFP

appli-Transparent Generic Framing Procedure

Several important high-speed LAN protocols use a Layer 1 block code inorder to communicate both wireless data and control information Themost common block code is the 8B/10B line code used for Gigabit Ethernet,

395

Data center A

Data center A

Data center B

Data center B

WDM conv WDM conv WDM conv

WDM conv WDM conv WDM conv

TGFP ingress

TGFP egress

ESCON GbE

Fibre Channel

ESCON GbE

Fibre Channel ESCON GbE

Fibre Channel ESCON GbE

ESCON, SBCON, Fibre Channel, FICON, and Infiniband, which havebecome increasingly important with the growing popularity of storage-area networks (SANs) Since both client wireless data bytes and datasource-to-sink control information are encoded into the 8B/10B codes,efficient transport of these protocols through a public transport networksuch as synchronous optical network/synchronous digital hierarchy(SONET/SDH) or the optical transport network (OTN) requires trans-porting both the wireless data and the 8B/10B control code information.The 8B/10B coding, however, adds a 25 percent wireless data bandwidthexpansion that is undesirable in the transport network.

The previously available protocols for LAN transport through SONET/SDH networks were asynchronous transfer mode (ATM) and packet overSONET/SDH (POS) ATM is relatively inefficient from a bandwidth uti-lization standpoint and typically requires a much more complex adaptationprocess than GFP POS requires terminating the client signal’s Layer 2 pro-tocol and remapping the signal into Point-to-Point Protocol (PPP) overHDLC, which suffers from a nondeterministic bandwidth expansion dis-cussed previously on bandwidth considerations Also, neither ATM norPOS supports the transparent transport of the 8B/10B control characters

In order to overcome the shortcomings of ATM and POS, GFP tion began in the American National Standards Institute (ANSI) accreditedT1X1 subcommittee, which chose to work with the International Telecom-munication Union—Telecommunication Standardization Sector (ITU-T) onthe final version of the standard, which has been published by the ITU-T.The transparent version of GFP has been optimized for transparently car-rying block-coded client signals (both the data and the 8B/10B controlcodes) with minimal latency This part of the chapter begins with adescription of the transparent GFP protocol, followed by some special con-siderations such as bandwidth, error control, and client management.Potential extensions to the transparent GFP protocol are then also brieflydiscussed

standardiza-Transparent GFP Description: General GFP Overview

The basic GFP frame structure is shown in Fig 16-8.4Protocols such asHDLC that rely on specific wireless data patterns for frame delimiting orcontrol information require a nondeterministic amount of bandwidthbecause of the need for additional escape bits or characters adjacent to thepayload strings or bytes that mimic these reserved characters Theamount of expansion is thus data pattern–dependent In the extreme case,

if the client payload data consist entirely of data emulating these reservedcharacters, byte-stuffed HDLC protocols like POS require nearly twice the

Chapter 16: Packet-over-SONET/SDH Specification

bandwidth to transmit the packet than if the payload did not contain suchcharacters GFP avoids this problem by using information in its core headerfor frame delimitation Specifically, the GFP core header consists of a two-octet-long field that specifies the length of the GFP frame’s payload area inoctets, and a cyclic redundancy check (CRC-16) error check code over thislength field The framer looks for a 32-bit pattern that has the proper zeroCRC remainder and then confirms that this is the correct frame alignment

by verifying that another valid 32-bit sequence exists immediately afterthe current frame ends, as specified by the length field Since no specialcharacters are used for framing, there are no forbidden payload valuesthat require escape characters

on the core header once frame alignment has been acquired

In frame-mapped GFP (GFP-F), a single client data frame [an IP packet

or Ethernet medium access control (MAC) frame] is mapped into a singleGFP frame For transparent GFP, however, a fixed number of client charac-ters are mapped into a GFP frame of predetermined length Hence, the payload length is typically variable for frame-mapped GFP and static fortransparent GFP One of the primary advantages of TGFP over GFP-F isthat TGFP supports the transparent transport of 8B/10B control characters

as well as wireless data characters In addition, GFP-F typically incurs the

397

Core header

Payload area

16-bit payload length indicator cHEC (CRC-16)

Client payload field

Payload headers (4-64 bytes)

Type (4 bytes) Extension (0-60 bytes)

Optional payload FCS (CRC-32)

Figure 16-8

GFP frame format

latency associated with buffering an entire client data frame at the ingress

to the GFP mapper As discussed next, TGFP requires only a few bytes ofmapper/demapper latency This lower latency is a critical issue for SANprotocols, which are very sensitive to transmission delay

impor-tant than bandwidth efficiency For example, if the client signal is lightlyloaded, GFP-F allows mapping the packets into a smaller transport chan-nel or potentially frame multiplexing them into a shared channel withGFP frames from other client signals Alternatively, GFP-F could makeuse of the link capacity adjustment scheme (G.7042) for handling clientsignals that experience temporary changes to their required bandwidth

Transparent GFP 64B/65B Block Coding

The 8B/10B line code maps the 28⫽ 256 possible data values into the

210⫽ 1024 value 10-bit code space such that the running number of onesand zeros transmitted on the line (the running disparity) remains bal-anced over very short intervals Twelve of the 10-bit codes are reservedfor use as control codes that may be used by the wireless data source tosignal control information to the wireless data sink The first step ofTGFP encoding in the source adaptation process is to decode the client8B/10B codes into control codes and 8-bit data values Eight of thesedecoded characters are then mapped into the 8 payload bytes of a64B/65B code The leading (flag) bit of the 64B/65B code indicateswhether there are any control codes present in that 64B/65B code (withflag ⫽ 1 indicating the presence of a control code) The 64B/65B blockstructure for various numbers of control codes is illustrated in Fig 16-9.4

Control codes are placed in the leading bytes of the 64B/65B block asillustrated in Fig 16-9 A control code byte consists of a bit to indicatewhether this byte contains the last control code in that 64B/65B block(⫽ 0 if it is the last), a 3-bit address (aaa−hhh) indicating the originallocation of that control code in the wireless client data stream relative tothe other characters mapped into that 64B/65B block, and a 4-bit code(Cn) representing the control code Since there are only 12 defined8B/10B control codes, 4 bits are adequate to represent them One of theremaining 4-bit codes is used to communicate that an illegal 8B/10Bcharacter has been received by the GFP source adaptation process sothat the GFP receiver can output an equivalent illegal 8B/10B character

to the client signal sink Figure 16-10 illustrates mapping of control andwireless data octets in the 64B/65B block.4

Aligning the 64B/65B payload bytes with the SONET/SDH/OTN load bytes simplifies parallel wireless data path implementations, in

addition to increasing the payload data observability within theSONET/SDH stream In order to achieve this alignment, a group ofeight 64B/65B codes are combined into a superblock The superblockstructure, as shown in Fig 16-11, takes the leading flag bits of the eightconstituent 64B/65B codes and groups them into a trailing byte followed

by a CRC-16 over the bits of that superblock.4CRC-16 is discussed ther in the section “Error Control Considerations,” below

fur-Transport Bandwidth Considerations

TGFP channel sizes are chosen to accommodate the wireless client datastream under worst-case clock tolerance conditions (for the slowest end ofthe transport clock and fastest end of the client clock tolerance) In thecase of SONET/SDH, while TGFP can be carried over contiguously con-catenated channels, it will typically be carried over virtually concatenatedsignals The concept of virtual concatenation is one in which multipleSONET synchronous payload envelopes (SPEs) are grouped together withSDH virtual containers (VCs) to form a higher-bandwidth pipe betweenthe endpoints of the virtually concatenated path

Octet # 000

D1 001 K1 010 D2 011 D3 100 D4 101 K2 110 D5 111 D6 Client byte

stream

Octet # 000

1.001.C1 L 1

001 0.101.C2

010 D1 011 D2 100 D3 101 D4 110 D5 111 D6 65B code

stream

Octet # 000

D1 001 K1 010 D2

011 Buffer underflow

100 D3 101 K2 110 D4 111 D5 Client byte

stream

Octet # 000

1.001.C1 L 1

001 0.011.P1

010 0.101.C2

011 D1 100 D2 101 D3 110 D4 111 D5 65B code

block: (a) with

con-trol and wireless data

bytes; (b) including

65B_PAD insertion

Chapter 16: Packet-over-SONET/SDH Specification

which greatly simplifies the provisioning and increases the flexibility ofvirtual concatenation

Another advantage of virtual concatenation is that it is transparent tointermediate nodes, with only the endpoints of the virtually concatenatedpath needing to be aware of its existence The nomenclature for indicating

a virtually concatenated signal is ⬍SPE/VC type⬎-Xv, where X indicatesthe number of SPEs/VCs that are being concatenated For example, STS-3c-7v is the virtual concatenation of seven STS-3c SPEs, which is equiva-lent to VC-4-7v for SDH Virtual concatenation is specified in the ITU-T,the ANSI, and the European Telecommunications Standards Institute(ETSI) Table 16-2 shows the minimum virtually concatenated channelsize that can be used for various TGFP clients.4

In practice, the SONET/SDH channel must be slightly larger thanthat needed to carry the GFP signal, a consequence of which is that theGFP mapper’s client signal ingress buffer will underflow There are twoways to handle this situation One approach is to buffer an entire TGFPframe’s worth of wireless client data characters prior to beginning thetransmission of that GFP frame This approach would increase the mapper

401

Leading bit

64B/65B code block

8 ⫻ 64B/65B blocks Code leadingbits

repositioned into a trailing byte

CRC-16

Core and payload headers

Chapter 16: Packet-over-SONET/SDH Specification

latency and buffer size A second approach, which was adopted for thestandard, is to use a dummy 64B/65B control code as a 65B_PAD charac-ter Whenever there is no client character available in the ingress buffer,the mapper will treat the situation the same as if a client control characterwere present and will insert the 4-bit 65B_PAD character Figure 16-10b

illustrates the insertion of a 65B_PAD character The demapper at theother end of the GFP link recognizes this character as a dummy pad andremoves it from the wireless data stream The result of using this65B_PAD character is that the mapper ingress buffer size is reduced toeffectively 8 bytes (the amount of data required to form a 64B/65B block)plus the number of bytes that can accumulate during the SONET/SDHoverhead and the GFP frame overhead bytes An 8-byte latency is alwaysrequired since the mapper cannot complete the 64B/65B block codinguntil it knows whether there are any control codes present in the eightcharacters that will make up that block

As discussed next, client management frames (CMFs) have been posed for GFP that would make use of this “spare” bandwidth for clientmanagement applications These CMFs would be up to 20 bytes long(including GFP encapsulation bytes) and, because they have lower prioritythan the wireless client data, would be allowed to be sent only when theingress buffer is nearly empty Support for these CMFs adds 20 bytes tothe ingress buffer requirements to accommodate the wireless data arrivingduring the transmission of a CMF

pro-Demapping of the TGFP signal entails the removal of the 65B_PADcharacters, and removal and interpretation of the interframe CMFswhen present Assuming that the egress of the client signal is done using

a constant-rate local clock, if the egress buffer becomes empty as a result

of reception of 65B_PAD characters and/or CMFs, interpacket fill wordsmust be inserted according to the client signal type rules

Error Control Considerations

The 8B/10B codes have built-in error detection capability since a singlebit error will always result in an illegal code The increased bandwidthefficiency gained by decoding the 8B/10B codes and remapping the datainto 64B/65B codes comes at the expense of much of this error detectioncapability There are four situations in which errors can cause significantproblems with 64B/65B codes The first and most serious problem results

if the leading flag bit of the 64B/65B code is received in error If the nal block contained control codes, these codes will be interpreted as wire-less data, and if the original block contained only data, some of thesebytes may be interpreted as control codes The number of data bytes thatare erroneously interpreted as control codes depends on the value of the

origi-403

first bit (the last control code indicator bit position) of the bytes andwhether the values of the location address bit positions contain increasingvalues (which would always be the case for a legal block) Wireless dataerroneously converted into control codes could cause the truncation of awireless client data frame, which in turn can cause error detection prob-lems for the wireless client data, since there is a possibility of the truncat-

ed wireless client data frame appearing to have a correct CRC value Asimilar situation occurs when control characters are present and the lastcontrol code indicator bit is affected by an error Also, errors in the controlcode location address will cause it to be placed in the wrong sequence bythe demapper, and errors in a 4-bit control code value will cause thedemapper to generate an incorrect control code Any error that results in

a spurious or incorrect control code has potentially serious consequences

It is these potential error problems that lead to the addition of a CRC-16

to each superblock The most reliable mechanism for error control is forthe demapper to discard all of the wireless data in a superblock in which

an error is detected The wireless data are discarded by having the per output 10B_ERROR 8B/10B codes for those clients that have definedsuch a code, or another illegal 8B/10B character for all of the characters inthat superblock

single-error correction

The payload area of the GFP frame is scrambled with a nous scrambler, and another error control issue concerns the interactionbetween the GFP payload scrambler and the superblock CRC-16 Tounderstand the issue here, it is helpful to first understand the rationaleand implementation behind the payload scrambler

self-synchro-The reasons for using a self-synchronized payload scrambling processare related to the physical properties of the transport medium and thedesire for robustness in public networks The line code used forSONET/SDH and OTN is non-return-to-zero (NRZ) (after the data havebeen passed through a SONET/SDH/OTN frame-synchronous scram-bler) For NRZ, the laser is turned on for the bit period to represent a 1and off to represent a 0 The advantage of the NRZ line code is its sim-plicity and bandwidth efficiency The disadvantage of NRZ, however, isthat the receiver clock and data recovery circuits can lose synchroniza-tion after a long string of either 0s or 1s The frame-synchronized scram-bler, which is reset at regular intervals by the SONET/SDH/OTN frame,

is adequate to defend against normally occurring user data patterns Itwould be possible, however, for a malicious user to choose a packet pay-load that is the same as the frame-synchronized scrambler sequence Ifthis packet lines up in the correct position in the transport frame, an

Chapter 16: Packet-over-SONET/SDH Specification

adequately long string of 0s or 1s can be generated to cause a loss ofsynchronization at the receiver The resulting loss of synchronizationwill take down the transport link, while the receiver attempts to recover,thus denying the link to other users in the meantime This problem wasoriginally discovered in ATM networks and is exacerbated by the longerframes used in POS or GFP In order to guard against such attacksATM, POS, and GFP use a self-synchronous payload scrambler to fur-ther randomize the wireless payload data This self-synchronous scram-

bler uses a polynomial of x43 ⫹ 1, which means that each bit of theATM/POS/GFP payload area is exclusive ORed with the scrambler out-put bit that preceded it by 43 bit positions, as shown in Fig 16-12.4Thedecoder’s descrambler reverses this process

In order to use the same payload scrambling technique for both mapped and transparent GFP, all of the GFP payload bits including theTGFP superblock CRCs must be scrambled As a result, the superblockCRC has to be calculated over the superblock payload bits prior to scram-bling and checked at the decoder after descrambling The drawback to aself-synchronous scrambler, however, is that each transmission errorresults in a pair of errors (43 bits apart here) in the descrambled data,which means that the superblock CRC must cope with this error multipli-cation It has been shown that a CRC will preserve its error detectioncapability in this situation as long as the scambler polynomial and theCRC generator polynomial have no common factors Unfortunately, all of

frame-the standard CRC-16 polynomials contain x⫹ 1 as a factor, which is also

a factor in the x43⫹ 1 (or any x n⫹ 1) scrambler polynomial Therefore, anew CRC generator polynomial was required that preserved the triple-error detecting capability (which is the maximum achievable over thisblock size) without having any common factors with the scrambler Inorder to perform single-error correction, the syndromes for single errorsand double errors spaced 43 bits apart must all be unique The code

selected for the superblock is x16⫹ x15⫹ x12⫹ x10⫹ x4⫹ x3⫹ x2⫹ x ⫹ 1,

which has both these desired properties, and hence retains its triple-errordetection and optional single-error correction capabilities in the presence

in

Dn

D2 D1

n

Data out Data

in

Figure 16-12

Payload

self-synchro-nous scrambler

Transparent GFP Client Management Frames

CMFs have the same structure as GFP wireless client data frames, butare denoted by the payload type code PTI ⫽ 100 in the GFP payloadheader Like GFP wireless client data frames, CMFs have a core header,

a payload header [both with 2-byte header error checking (HEC)], and

an optional 32-bit FCS The total CMF payload size in TGFP is mended to be no greater than 8 bytes

recom-Assuming an 8-byte payload area along with the 8 total bytes for themandatory core and type headers, the payload efficiency will be 50 per-cent Use of FCS and especially extension headers will greatly reducethe efficiency of the CMFs

As previously noted, there is some residual “spare” bandwidth in theSONET/SDH channel for each of the client signal mappings As shown

in Table 16-2, the amount of this spare bandwidth depends on the ciency of the mapping, which in turn is partially a function of the num-ber of superblocks used in each GFP frame The residual bandwidth can

effi-be used as a client management overhead channel for client ment functions, as described in this part of the chapter CMFs are alsoused for downstream indication of client signal fail

manage-Remote Management If both ends of the GFP link are owned by thesame carrier and the intervening SONET/SDH/OTN network is owned byanother operator, the potential exists for sending GFP-specific provision-ing commands using CMFs It is not uncommon for interexchange carri-ers (IECs) to provide the customer premises equipment (CPE) and rely

on a local exchange carrier (LEC) to provide the connection between theCPE and the IEC network (see Fig 16-13).4Ideally, the IEC would like tomanage the CPE as part of its own network, which frees the customerfrom having to manage the equipment and allows the IEC a potential rev-enue source from providing the management service Normally, manage-ment information is communicated through a wireless SONET/SDH sectiondata communication channel (SDCC) In order to prevent unwanted con-trol access, however, carriers do not allow SDCC wireless data to cross thenetwork interfaces into their networks; hence, there is currently no wayfor the IEC to exchange management communications with the CPEthrough the intervening LEC network TGFP CMFs, however, provide amechanism to tunnel the SDCC information through the intervening net-work Table 16-2 shows the maximum amount of payload capacity thatcan be derived from the CMFs for SDCC tunneling or other operations,administration, and maintenance (OAM) applications, with the assump-tions stated in the table notes and assuming a 20-byte CMF with an 8-bytepayload field For all client signal types, there is adequate bandwidthavailable to carry a 192-kbps SDCC channel

Chapter 16: Packet-over-SONET/SDH Specification

Conclusion

This chapter introduced several emerging techniques currently underdevelopment for next-generation SONET/SDH systems On the basis ofthese new techniques, the chapter elaborated on new SONET/SDHtransport services likely to become reality within a few years Data overSONET/SDH, using GFP, virtual concatenation, and LCAS, is likely tobecome the dominant transport method over SONET/SDH transportnetworks

Looking ahead, flexible transport services, combined with virtuallyunlimited bandwidth availability brought by WDM transport techniques,will ensure that sophisticated and bandwidth-hungry Internet applica-tions of the future can be deployed These yet to be seen applications willlikely change computer and human communications in a revolutionaryand unprecedented way

Finally, transparent GFP provides an efficient mechanism for ping constant-bit-rate block-coded wireless data signals across aSONET/SDH network or OTN Performing the mapping on a client char-acter basis rather than a client frame basis significantly reduces thetransport latency to a fixed number of bytes rather than a whole clientframe, which is a critical issue for SAN protocols including Gigabit Eth-ernet The translation of client block codes into more efficient 64B/65Bmapping provides significant bandwidth efficiency increase while thesuperblock structure provides robustness Transparent GFP also allowsincreased performance monitoring capability for the Transport layer, andthe ability to tunnel SDCC management information through an inter-vening network provides a powerful extension to network providers’capabilities

map-407

IEC’s NMS

IEC network

LEC network

GFP overhead

Customer premises

SONET terminal with GFP

VCG channel GFP client data

Figure 16-13

An SDCC tunneling

application example

with transparent GFP

3 Dirceu Cavendish, Kurenai Murakami, Su-Hun Yun, Osamu Matsuda,and Motoo Nishihara, “New Transport Services for Next-Generation

SONET/SDH Systems,” IEEE Communications Magazine, 445 Hoes

Lane, Piscataway, NJ 08855, 2002

4 Steven S Gorshe and Trevor Wilson, “Transparent Generic FramingProcedure (GFP): A Protocol for Efficient Transport of Block-Coded Data

through SONET/SDH Networks,” IEEE Communications Magazine,

445 Hoes Lane, Piscataway, NJ 08855, 2002

5 John R Vacca, i-mode Crash Course, McGraw-Hill, 2001.

6 John R Vacca, High-Speed Cisco Networks: Planning, Design, and Implementation, CRC Press, 2002.

Wireless Data

Access Implementation

The explosive growth of both the wireless industry and the Internet iscreating a huge market opportunity for the implementation of wireless dataaccess methods Limited Internet access, at very low speeds, is alreadyavailable as an enhancement to some existing cellular systems However,those systems were designed with the purpose of providing voice servicesand (at most) short messaging, but not fast data transfers In fact, asshown in this chapter, traditional wireless data technologies are notvery well suited to meet the demanding requirements of providing veryhigh data rates with the ubiquity, mobility, and portability characteristic

of cellular systems Increased use of antenna arrays appears to be theonly means of enabling the types of data rates and capacities needed forwireless data Internet and multimedia services While the deployment

of base station arrays is becoming universal, it is really the ous deployment of base station and terminal arrays that can unleashunprecedented levels of performance by opening up multiple spatial sig-naling dimensions

simultane-Using Antenna Arrays: Lifting the Limits on High-Speed

Wireless Data Access

As previously mentioned, the explosive growth of both the wireless dataindustry and the Internet is creating a huge market opportunity for wire-less data access Limited Internet access at low speeds (a few tens of kilo-bits per second at most) is already available as an enhancement to somesecond-generation (2G) cellular systems However, those systems wereoriginally designed with the sole purpose of providing voice services and,

at most, short messaging, but not fast data transfers Third-generation(3G) mobile wireless data systems,3currently under development, willoffer true packet access at significantly higher speeds Theoretically, userdata rates as high as 2 Mbps will be supported in certain environments,although recent studies have shown that approaching those rates might

be feasible only under extremely favorable conditions—in the vicinity of abase station and with no other users competing for bandwidth

In fact, as will be argued in this part of the chapter, traditional less data technologies are not particularly well suited to meet theextremely demanding requirements of providing the very high datarates and low cost associated with wired access, and the ubiquity, mobil-ity, and portability characteristic of cellular systems Some fundamentalbarriers, related to the nature of the radio channel as well as to limitedbandwidth availability at the frequencies of interest, stand in the way

wire-Chapter 17: Wireless Data Access Implementation Methods

As a result, the cost per bit in wireless data is still high and not ishing fast enough In contrast, the wired world is already providingbasically free bits, which has accustomed an entire generation of Internetusers to accessing huge volumes of information at very high speeds andnegligible cost This part of the chapter establishes practical limits on thewireless data rates that can be supported by a wireless data access sys-tem with a typical range of parameters, and it shows how those limitscan be lifted by using a combination of transmit and receive antennaarrays with powerful space-time processing techniques (The Glossarydefines many technical terms, abbreviations, and acronyms used in thebook.)

dimin-Fundamental Limitations in Wireless Data Access

Ever since the dawn of the information age, capacity has been the cipal metric used to assess the value of a communication system How-ever, several definitions of capacity exist Link capacity or user capacity

prin-is used here to signify the highest data rate at which reliable cation is possible between a transmitter and a receiver At the sametime, system capacity is used here to indicate the total throughput (sum

communi-of user data rates) within a cell or sector System capacity can be verted into area capacity simply via normalizing by the cell size Sinceexisting cellular systems were devised almost exclusively for telephony,user data rates were low and had minimal variability In fact, sourcerates were purposefully reduced to the minimum level necessary to sup-port a highly compressed voice call and implicitly traded for additionalusers Therefore, systems were designed to accommodate a large num-ber of low-data-rate users With the emergence of wireless data services,many of these concepts are becoming obsolete User wireless data ratesare increasingly variable and heterogeneous The value of a system is nolonger defined only by how many users it can support, but also by itsability to provide high peak rates to individual users as needed—inother words, by its ability to concentrate large amounts of capacity atvery localized spots Thus, in the age of wireless data, user data ratesurges again as an important metric

con-Because of the logarithmic relationship between the capacity of awireless data link and the signal-to-interference-and-noise ratio (SINR)

at the receiver, trying to increase the wireless data rate by simply mitting more power is extremely costly Furthermore, it is futile in thecontext of a dense interference-limited cellular system, wherein anincrease in everybody’s transmit power scales up both the desired signals

trans-as well trans-as their mutual interference, yielding no net benefit Therefore,

411

power increases are useless once a system has become limited in essence

by its own interference Furthermore, since mature systems designed forhigh capacity tend to be interference-limited, it is power itself (in theform of interference) that ultimately limits their performance As aresult, power must be carefully controlled and allocated to enable thecoexistence of multiple, geographically dispersed users operating in var-ious conditions Hence, power control has been a topic of very activeresearch for many years

Increasing the signal bandwidth (along with the power) is a moreeffective way of augmenting the wireless data rate However, radio spec-trum is a scarce and very expensive resource at the frequencies of inter-est, where propagation conditions are favorable Moreover, increasingthe signal width beyond the coherence bandwidth of the wireless datachannel results in frequency selectivity Although well-established tech-niques such as equalization and orthogonal frequency-division multi-plexing can address this issue, their complexity grows rapidly with thesignal bandwidth Altogether, it is imperative that every unit of band-width be utilized as efficiently as possible Consequently, spectral effi-ciency (defined as the capacity per unit bandwidth) has become anotherkey metric by which wireless data systems are measured In order toimprove it, multiple-access methods (originally rather conservative intheir design) have evolved toward much more sophisticated schemes In thecontext of frequency-division multiple access (FDMA) and time-divisionmultiple access (TDMA), this evolutionary path has led to advancedforms of dynamic channel assignment that enable adaptive and muchmore aggressive frequency reuse In the context of code-division multipleaccess (CDMA), it has led to a variety of multiuser detection and inter-ference cancellation techniques

Models and Assumptions

This analysis is conducted in the 2-GHz frequency range, which is where3G systems will initially be deployed This is a favorable band from apropagation standpoint Also, and again in line with the 3G framework,

the available bandwidth is assumed to be B⫽ 5 MHz However, for plicity, frequency selectivity is ignored here, with the argument that itcan be dealt with by using the techniques mentioned earlier and theirextension to the realm of antenna arrays and BLAST

sim-The downlink has the most stringent demands for wireless dataapplications However, a similar analysis could be applied to the uplink,although with much tighter transmit power constraints A cellular sys-tem with fairly large cells is assumed, with every cell partitioned into120° sectors

Chapter 17: Wireless Data Access Implementation Methods

The propagation scenario portrayed in Fig 17-1 is based on the tence of an area of local scattering around each terminal.1Little or nolocal scattering is presumed around the base stations From the perspec-tive of a base station, the angular distribution of power that gets scat-tered to every terminal is characterized by its root-mean-square width,commonly referred to as angle spread Typical values for the angle spread

exis-at a base stexis-ation are in the range of 1 to 10°, depending on the ment and range The antennas composing a base station array can beoperated coherently if they are closely spaced, or decorrelated by spacingthem sufficiently apart At a terminal buried in clutter, on the other hand,angle spreads tend to be very large (possibly as large as 360°) and thus,large uncorrelation among its antennas is basically guaranteed

environ-M is used here to denote the number of antennas within every base tion sector and N to indicate the number of antennas at every terminal.

sta-The channel responses from every sector antenna to every terminal

antenna are assembled into an N ⫻ M channel matrix H ⫽ {h nm} ing frequency selectivity, the entries of H are complex gaussian scalars(Rayleigh distributed in amplitude), whose local average path gain has

Ignor-a rIgnor-ange-dependent component Ignor-and Ignor-a shIgnor-adow fIgnor-ading component Therange-dependent component is modeled here using the well-establishedCOST231 model The shadow fading is taken to be log-normally distrib-uted with an 8-dB standard deviation The correlation among the entries

of H is determined by the antenna spacing and angle spread Antennaswithin a terminal are assumed fully uncorrelated, whereas those within abase station sector are assumed to be either uncorrelated (for sufficientlylarge spacing) or fully correlated (for close spacing and coherent opera-tion) Therefore, the rows of H are always independent, whereas thecolumns can be either linearly dependent or also independent

413

Figure 17-1

A propagation

sce-nario with local

scat-tering around the

terminal spanning

a significant angle

spread at the base

station

While power control proved to be an essential ingredient in voice tems, where source rate variability was minimal, in wireless data systemsrate adaptation becomes not only an attractive complement, but even a fullalternative to power control Hence, this part of the chapter restricts itself

sys-to the case where the sys-total power per user is held constant while the less data rate is being adapted

wire-Furthermore, an open-loop architectures is used here—whereintransmitters do not have access to the instantaneous state of the chan-nel Only long-term information (information that varies slowly withrespect to the fading rate) is available to the transmitters In its originalform, BLAST has no need for instantaneous channel information at thetransmitter However, more elaborate closed-loop forms of BLAST havebeen devised in order to exploit that information in those cases when itmay be available, such as when a fast feedback link is available or whentime-division duplexing is employed It is assumed (in all cases) that thechannel matrix H is known perfectly at the receiver

acqui-sition algorithms

Single-User Wireless-Data-Rate Limits

Let us first consider an isolated single-user link limited only by thermalnoise Within the context of a real system, this would correspond to anextreme case wherein the entire system bandwidth is allocated to an indi-vidual user Furthermore, it would require that no other users be activeanywhere in the system or that their interference be perfectly suppressed.Clearly, these are unrealistic conditions; thus, the single-user analysisprovides simply an upper bound, only a fraction of which is attainable.Also, since for a system to be interference-limited, it is necessary that the

signal-to-noise ratio (S/N) be large enough that the noise level is much

lower than the interference, this analysis also determines what cell sizescan be supported in interference-limited conditions

Steered Directive Array The use of an array is first considered here

at the base station only, with M closely spaced antennas operating

coher-ently Since an estimate of the directional location of the terminal can beusually derived from the uplink, the use of directive array algorithms hasbeen regarded as an attractive option for enhancing the performance ofexisting 2G systems

Each individual base station antenna has a gain of 15 dBi The nal is equipped with a single omnidirectional antenna Under these con-ditions, and assuming the beam synthesized at the base is properly

termi-steered toward the terminal, as M grows, the array becomes more

direc-Chapter 17: Wireless Data Access Implementation Methods

tive; thus, more precise information about the directional location of theterminal is needed in order to fully illuminate its local scattering area.Also, since no further directional gain can be realized beyond the point

at which the beamwidth becomes smaller than the angle spread, the size

of a directive array has a fundamental bound imposed by the

environ-ment Now, set M⫽ 8, at which point the beamwidth falls below 10°, asthe maximum number of 15-dBi antennas that can be aggregated.The 90 percent single-user capacities corresponding to this scenario

are presented in Fig 17-2 with the transmit power set to P

T⫽ 10 W.1

Because of the logarithmic relationship between rate and power, the use

of a base station directive array offers very limited improvement interms of single-user data rate

Transmit Diversity An alternative strategy, also based on the ment of base station arrays only, which has already been incorporatedinto the 3G roadmap, is that of transmit diversity In this case, the basestation antennas must be spaced sufficiently far apart so that their sig-nals are basically uncorrelated The corresponding single-user results, asshown in Fig 17-3, are nonetheless similar to their directive array coun-terparts.1However, in this case, there is no fundamental bound to the

deploy-size of the array; there is little advantage in increasing it beyond M⫽ 3

to 4 because of the diminishing returns associated with adding additionaldiversity branches to an already diverse link Furthermore, in a truewideband system, frequency diversity would further reduce the benefits

Directive array (M antennas) at base station

Single antenna at terminal

8 4

Multiple-Transmit Multiple-Receive Antenna Architectures

Now, let’s look at architectures with both transmit and receive arrays

As in the transmit diversity case, base station antennas must be spacedapart for proper decorrelation In addition, the terminal must beequipped with its own array Also, as in the diversity case, no informa-tion about the directional location of the terminal is required In order toavoid cluttering the results with an excessive number of parameters,scale the size of both the base station and the terminal arrays simulta-

neously; that is, set M ⫽ N The capacities are depicted in Fig 17-4.1Forcompleteness, the transmit diversity curves of Fig 17-3 are also shown.Notice the extraordinary growth in attainable data rates unleashed bythe additional signaling dimensions provided by the combined use of

transmit and receive arrays With only M ⫽ N ⫽ 8 antennas, the

wire-less single-user data rate can be increased by an order of magnitude thermore, the growth does not saturate as long as additional uncorrelatedantennas can be incorporated into the arrays

Fur-Wireless-Data-Rate Limits within

a Cellular System

In this part of the chapter, let’s extend the analysis in order to reevaluatethe wireless user data-rate limits in much more realistic conditions Tothat end, let’s incorporate an entire cellular system into the analysis

per-cent of locations

ver-sus range with

Chapter 17: Wireless Data Access Implementation Methods

Most emerging wireless data-oriented systems feature time-multiplexeddownlink channels, certainly those evolving from TDMA, but also thoseevolving from CDMA With that, same-cell users are ensured to bemutually orthogonal; thus, the interference arises exclusively from othercells Accordingly, a time-multiplexed multicell system is consideredwith base stations placed on a hexagonal grid Users are uniformly dis-tributed and connected to the sector from which they receive thestrongest signal To further mimic actual 3G data systems, rate adapta-tion with no power control is assumed Transmit signals are assumedgaussian, which maximizes capacity as long as no multiuser detectionacross cells is attempted Altogether, the results presented in this part ofthe chapter can be considered upper bounds for a 5-MHz data-oriented3G system

The results correspond to Monte Carlo simulations conducted on a 19-celluniverse: a central cell, wherein statistics are collected, surrounded by tworings of interfering cells The cell size is scaled to ensure that the system isbasically interference-limited The simulation parameters are summa-rized, for convenience, in Table 17-1.1

Figure 17-5 displays cumulative distributions of system capacity (inmegabits per second per sector) over all locations with transmit arraysonly, as well as with transmit and receive arrays.1These curves can also beinterpreted as user peak rates, that is, wireless user data rates (inmegabits per second) when the entire capacity of every sector is allocated

to an individual user With transmit arrays only, the benefit appears

BLAST DIV

Figure 17-4

Single-user data rate

supported in 90

per-cent of locations

versus range with

transmit and receive

arrays M is the

num-ber of 15-dBi

anten-nas at the base

station as well as the

significant only in the lower tail of the distribution, corresponding to users

in the most detrimental locations The improvements in average and peaksystem capacities are negligible Moreover, the gains saturate rapidly asadditional transmit antennas are added With frequency diversity takeninto account, those gains would be reduced even further The combined use

of transmit and receive arrays, on the other hand, dramatically shifts thecurves, offering multifold improvements in wireless data rate at all levels

NOTE Without receive arrays, the peak wireless data rate that can be

sup-ported in 90 percent of the system locations (with a single user per sector) isonly on the order of 500 kbps with no transmit diversity and just over 1 Mbpstherewith Moreover, these figures correspond to absolute upper bounds

With modulation excess bandwidth, training overhead, imperfectchannel estimation, realistic coding schemes, and other impairments,only a fraction of these bounds can be actually realized Without receivearrays, user rates on the order of several megabits per second can besupported only within a restricted portion of the coverage area andwhen no other users compete for bandwidth within the same sector.Finally, the broadband wireless access industry, which provides high-rate network connections to stationary sites, has matured to the point atwhich it now has a standard for second-generation wireless metropolitan-area networks IEEE Standard 802.16, with its WirelessMAN air inter-face, sets the stage for widespread and effective deployments worldwide.The next part of the chapter is an overview of the technical mediumaccess control and Physical layer features of this new standard

MultiplexingTime Division

Base station antennas 120° perfect sectorization Terminal antennas Omnidirectional Frequency reuse Universal Propagation exponent 3.5 Log-normal shadowing 8 dB Fading Rayleigh (independent per antenna)

Chapter 17: Wireless Data Access Implementation Methods

WirelessMAN: Air Interface for Broadband Wireless Access

IEEE Standard 802.16-2001, completed in October 2001 and published

on April 8, 2002, defines the WirelessMAN air interface specification forwireless data metropolitan-area networks (MANs) The completion ofthis standard heralds the entry of broadband wireless data access as amajor new tool in the effort to link homes and businesses to core telecom-munications networks worldwide

As currently defined through IEEE Standard 802.16, a wireless dataMAN provides network access to buildings through exterior antennas com-municating with central radio base stations (BSs) The wireless data MANoffers an alternative to cabled access networks, such as fiber-optic links,coaxial systems using cable modems, and digital subscriber line (DSL)links Because wireless data systems have the capacity to address broadgeographic areas without the costly infrastructure development required

in deploying cable links to individual sites, the technology may prove lessexpensive to deploy and may lead to more ubiquitous broadband access.Such systems have been in use for several years, but the development ofthe new standard marks the maturation of the industry and forms thebasis of new industry success with second-generation equipment

In this scenario, with WirelessMAN technology bringing the network

to a building, users inside the building will connect to it with tional in-building networks such as, for data, Ethernet (IEEE Standard

conven-419

1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Transmit diversity (single antenna

at terminals)

BLAST

DIV

BLAST BLAST

Figure 17-5

Cumulative

distribu-tions of system

capac-ity with transmit

arrays as well as with

transmit and receive

arrays M is the

num-ber of antennas per

array; system