TCP over Wireless Networks

(VTC2000-Fall Tutorial)

Farooq Khan and Sanjiv Nanda

Bell Labs - Lucent Technologies
Holmdel, NJ 07733, USA

September 25, 2000

 Why TCP?
Explosive growth of internet.

Web browsing (HTTP) contribute to more than 70%

of the traffic in the internet.

HTTP and some other internet applications use TCP

as the transport mechanism.

TCP is tuned to fixed networks (low delay, low error

rate).

Scalability of the internet is enabled by TCP flow

control mechanisms.
 Why TCP over Wireless?

Wireless networks that enable high-speed data

transfer (3G) are now being introduced.

Web browsing (using TCP) is expected to be the

dominant application in wireless data networks.

Wireless networks exhibit high delays and high error

rates.

This causes some internet protocols, especially TCP,

to perform poorly.
 How data is different from voice?

HTTP/TCP

Loss, Delay Variation When and how much

Air Interface (RLP,

MAC, PHY)
 Simplified Protocol Stack

Host

HTTP/TCP HTTP/TCP

internet

RLP/MAC/PHY RLP/MAC/PHY

BSC/RNC MS

HTTP/TCP Internet protocols

RLP/MAC/PHY Radio network protocols

 HTTP Request to Radio Frames
Receiver Sender
www.yahoo.com

HTTP GET request

HTTP www.yahoo.com http://www.yahoo.com

TCP HTTP www.yahoo.com

IP TCP HTTP www.yahoo.com

Segmentation
RLP RLP

Coding and interleaving etc.

Radio frames
 Tutorial Outline

Introduction

RLP TCP HTTP

TCP over Wireless

Performance Issues

TCP Performance
 Radio Link Protocol

Radio Link Protocol Provides:

Segmentation and Re-assembly
In sequence delivery of data
Automatic Ratransmission reQuest (ARQ)

Examples:
Radio Link Control (RLC) in UMTS and GPRS/EDGE.
Radio Link Protocol (RLP) in cdma2000 (IS-2000).
 IS-2000 RLP
NAK-Based

RLP Type 3 supports 2 rounds of NAKs.

Two NAKs in the first round and three NAKs in the second
round.

In RLP Type 3 enhanced, the number of rounds and

NAKs/retransmissions sent in each round are configurable.

The configuration parameters are determined during RLP

set up process through negotiation between the mobile
terminal and the network.
 RLP Type 3

TX

1 2 3 NAKs NAKs

2 2

2 2 2
RX

RTX Timer Abort Timer

At the expiry of abort timer, the RLP receiver will deliver

frames 1 and 3 to higher layer.
 Transmission Control Protocol (TCP)

TCP is a connection oriented transport layer protocol that

provides:

Reliable byte stream transport

In sequence delivery of data
reorders out-of-order data
discards duplicate data
End to end flow control
Calculates and verifies a mandatory checksum
Interface to the applications (e.g., HTTP, FTP, Email and
Telnet etc.)
 TCP Header

16-bit source port number 16-bit destination port

32-bit sequence number

32-bit acknowledgment number 20 bytes

4-bit Reserved 16-bit window size
HLEN

16-bit TCP checksum 16-bit urgent pointer

URGACKPSHRSTSYN FIN
 TCP Connection Management

TX
SYN(J) Data flow ACK(M+1) FIN(N)
ACK(K+1)

SYN(K) ACK(J+1) FIN(M) ACK(N+1)

RX

J
K+1 20 bytes

URG ACK PSH RST SYN FIN [SYN(J), ACK(K+1)]

 Loss Detection

TCP Data and ACKs can get lost due to congestion

or transmission errors.

Two different ways to detect loss:

Retransmission timeout
Duplicate ACKs
 TCP Timeout

TCP sets a timeout, when it sends data.

If the data is NOT acknowledged when the timeout

expires

TCP retransmits the data.

 Exponential Back-off

RTO=T RTO=2T RTO=4T

TX

RX

RTO = mean RTT+ 4(mean deviation of RTT)

 Duplicate ACKs

After receiving three duplicate ACK,

TCP sender assumes that the segment is lost and,

Perform a retransmission
RTX on Duplicate ACKs

TX RX
K
K+1
K+1
K+2
K+2
K+3
K+4 K+2
K+5 K+2 3 dupacks
K+6 K+2
K+7 K+2
K+2 K+2
K+8
 TCP flow Control

The congestion window (cwnd) is flow control

imposed by the sender.

While the advertised window is flow control imposed

by the receiver.

The former is based on the sender's assessment of

perceived network congestion; the latter is related to
the amount of available buffer space at the receiver
for this connection.

Current window = min(cwnd, advertised win).

 Window Based Flow Control

Can send ASAP

Window size

1 2 3 4 5 6 7 8 9 10

ACKs received Can't send until

window moves
Sent, not ACKed
 TCP Slow Start

Initially, cwnd=MSS;

Each time an ACK is received, cwnd cwnd+MSS;

This provides an exponential growth, although it is

not exactly exponential because the receiver may
delay its ACKs, typically sending one ACK for every
two segments that it receives.

Maximum Segment Size (MSS) is the largest chunk of data that TCP
will send to the other end. When a connection is established, each end
can announce its MSS.
 Congestion Avoidance

Congestion avoidance and slow start require that two

variables be maintained for each connection: a
congestion window, cwnd, and a slow start threshold
size, ssthresh.

Initially, cwnd MSS, ssthresh=65KB.

Current win min (cwnd, advertised window)

 Congestion Avoidance

When congestion occurs (indicated by a timeout or

the reception of duplicate ACKs), ssthresh max
[2*MSS, 1/2(current window size)]

Additionally, if the congestion is indicated by a

timeout, cwnd MSS (i.e., slow start).

When new data is acknowledged by the other end,

increase cwnd, but the way it increases depends on
whether TCP is performing slow start (cwnd
ssthresh) or congestion avoidance (cwnd > ssthresh).
 Congestion Avoidance Example

Current Advertised cwnd ssthresh

Window window

t 1.5KB 65KB 1.50KB 65KB

t+ 65KB 65KB 75.0KB 65KB
RTX due to 65KB 65KB 75.0KB 32.5KB
dupacks
RTX due to 1.50KB 65KB 1.50KB 32.5KB
timeout
 Congestion Avoidance

In congestion avoidance, each time an ACK is received,

cwnd MSS*MSS/cwnd. For example, MSS=1500 bytes,
current cwnd=6000 bytes, the new cwnd = 6375 bytes

This is a linear growth of cwnd, compared to slow start's

exponential growth.

The increase in cwnd should be at most one MSS each

RTT (regardless how many ACKs are received in that
RTT), whereas slow start increments cwnd by the number
of ACKs received in an RTT.
 SS and Congestion Avoidance

25
20
cwnd (segments)

15 ssthresh
10
5
0
0 1 2 3 4 5 6 7 8
RTT
 HyperText Transfer Protocol (HTTP)

HTTP is a simple protocol that forms the basis for the

Web.

The HTTP client (web browser) establishes a TCP

connection to the HTTP server

Issues a request, and reads back the servers

response

The server closes the TCP connection at the end of its

response.
 Web Browsing (HTTP/1.0)
Web browser Web server

TCP TCP

Internet

The biggest performance problem associated with HTTP is its

use of one TCP connection per web object (file).
 HTTP/1.0 (Continued)

HTTP/1.0 opens and closes a new TCP connection

for each web object transfer.

Most Web objects are small (1-5KB).

High fraction of packets are simply TCP control

packets used to open and close a connection.
 HTTP/1.0 (Continued)

Due to TCP slow start, most objects are transferred

before their TCP connection completes the slow start
algorithm.

Most HTTP/1.0 operations use TCP at its least

efficient.

Major problems due to resulting congestion and

unnecessary overhead.
 HTTP/1.0 (Continued)
Web object (A) transfer time

Completion of index.html
A Packet Bursts

B
4 parallel TCP connections

E
Completion of E and start of F

F
Web page transfer time
 HTTP/1.0 (Continued)

First TCP connection (A) is for the home page

(GET /).

This HTML document refers to 5 GIF images (B

through F).

As soon as this home page is received by the client,

four TCP connections are opened in parallel for the
first four images (B through E).

As soon as one of these connections terminates,

another connection is immediately established to
fetch the next and last image F.
 HTTP/1.1

HTTP/1.1 leaves the TCP connection open between

consecutive operations.

This technique is called "persistent connections,"

which both avoids the costs of multiple opens and
closes and reduces the impact of slow start.

Persistent connections are more efficient than the

current practice of running multiple short TCP
connections in parallel.
 HTTP/1.1 (Continued)

By leaving the TCP connection open between

requests, many packets can be avoided

While avoiding multiple RTTs due to TCP slow start.

 Related IETF Working Groups

IETF HTTP (HyperText Transfer Protocol) Working

Group
http://www.ietf.org/html.charters/http-charter.html

IETF Transport Area (TCP) Working Group

http://www.ietf.org/html.charters/tsvwg-charter.html

IETF PILC (Performance Implications of Link

Characteristics) Working Group
http://www.ietf.org/html.charters/pilc-charter.html
 Air Interface (AI) Characteristics

Wireless networks are characterized by:

High Latency (over-the-air and in the access network)

Variable Delays (e.g., due to link layer retransmissions)

High Error Rate (e.g., due to fading etc.)

Variable Data Rate (due to varying channel conditions)

Handoffs (due to mobility)

 High Latency (1)

Wireless System Delay (RTT) Bandwidth Delay*Bandwidth Product

(ms) (Kb/s) (Bytes)
W-LAN 3 1500 563
GSM, IS-95 500 9.6 600
Richochet 500 128.0 8KB
3G Wireless 200 384 9600 > 8KB

The delay bandwidth product for 3G wireless is larger

than the default 8KB buffer space used by many TCP
implementations.
 High Latency (2)
Delay=100ms, Rate=384Kb/s

12ms
Data
TCP
40ms
sender 576B
Wasted BW TCP
Max receiver
buffer
space
=8KB

12ms ACKs
Delay=100ms, Rate=384Kb/s

Maximum number of outstanding segments = Floor(8KB/576B)=13

 High Latency (3)

A buffer space (window size) smaller than the delay

bandwidth product result in wasted bandwidth due to
inefficient pipe filling.

The maximum achievable throughput is limited to

(window size)/RTT.
 Variable Delays

TCP RTO = mean RTT+4 (mean deviation of RTT)

Host

HTTP/TCP HTTP/TCP

internet

RLP/MAC/PHY Errors RLP/MAC/PHY

BSC/RNC MS
RLP Retransmissions

Large variations in the RTT (due to RLP retransmissions) increase the RTO
 High Error Rate

Loss due to buffer over-flow Loss due to RLP residual errors

in the internet routers (after max number of RTX attempts)

Router
Host RNC Errors MS

TCP congestion control is tuned to networks that lose

traffic primarily because of congestion and buffer
exhaustion.
Interact badly with connections that lose data due to
transmission errors.
 Variable Data Rate

Variable data rates introduce transmission time

variations, leading to RTT variations.

Large variations in the RTT increase the RTO.

Increased RTO increases the overall data transfer

times (e.g., web page transfer times) due to slow
response to recover from lost data.

Data rate variations (high data rate to low data rate)

in the start of a connection can lead to duplicate
transmissions from TCP.
 Handoffs

Variations in RTT due to delay associated with

Handoffs.

TCP timeouts (and duplicate transmissions) due to

sudden increase in delays seen by TCP (RTO cannot
be adjusted so fast).
 Potential Enhancements to TCP
Fast Rtx. and Fast Recovery

Slow Start Proposals

Delayed Duplicate Acks

Selective Acks

Split TCP and PEPs

For a complete list of enhancements, see:

draft-montenegro-pilc-ltn-01.txt.
 Fast Retransmit and Recovery

When the third duplicate ACK in a row is received, ssthresh

max (2*MSS, cwnd/2)

Retransmit the missing segment, cwnd ssthresh +

3*MSS.

This inflates the congestion window by the number of

segments that have left the network and which the other
end has cached.
 Fast Retransmit and Recovery

Each time another duplicate ACK arrives, cwnd

cwnd +MSS.

This inflates the congestion window for the additional

segment that has left the network.

Transmit a packet, if allowed by the new value of

cwnd.
 Fast Retransmit and Recovery

When the next ACK arrives that acknowledges new data,

cwnd ssthresh.

This ACK should be the acknowledgment of the missing

retransmission, one round-trip time after the retransmission.

Additionally, this ACK should acknowledge all the

intermediate segments sent between the lost packet and
the receipt of the first duplicate ACK.

This step is congestion avoidance, since TCP is down to

one-half the rate it was at when the packet was lost.
 Fast Retransmit and Recovery

The fast retransmit algorithm first appeared in the

4.3BSD Tahoe release, and it was followed by slow
start. The fast recovery algorithm appeared in the
4.3BSD Reno release.
 Fast RTX and Fast Recovery
TX RX
K
K+1
K+1
K+2
K+2
K+3
K+4 K+2
K+5 K+2 3 dupacks
K+6 K+2
K+7 K+2
Fast RTX K+2 K+2
K+8 K+8
Fast Recovery K+9 K+9
K+10
 Slow Start Proposals

Slow start dominates the network response seen by

interactive users at the beginning of a TCP
connection.

Wireless links present long delays

Stability of the internet is important.

The slow start proposals should not adversely affect

internet congestion levels.
 Slow Start Proposals

Slow Start Proposals

Larger Initial window Handling ACKs Terminating Slow

Start

Ack counting Ack-every-segment

 Larger Initial Window

Initial win = min (4*MSS, max (2*MSS, 4380 bytes))

Equivalently,
If (MSS <= 1095 bytes) then win <= 4 * MSS
If (1095 bytes < MSS < 2190 bytes) then win <=4380 bytes
If (2190 bytes <= MSS) then win <= 2 * MSS
 Larger Initial Window

25
cwnd (segments)

20
ssthresh
15 IW=1
10 IW=4
5
0
0 1 2 3 4 5 6 7 8
RTT

Beneficial for short-lived TCP connections or links with long RTT

 ACK Counting

Make each ACK count to its fullest by growing the

window based on the data being acknowledged (byte
counting).

May cause bursts that lead to congestion.

Limit the window growth to 2 segments.

 ACK-every-segment
Keep a constant stream of ACKs coming back by turning off
delayed ACKs during slow start.

Should be limited to slow start only in order to avoid penalizing

asymmetric-bandwidth configurations

200-500ms

K+1
K+1
K+2
 Terminating Slow Start

cwnd=ssthresh (instead of cwnd=MSS).

Theoretically, an optimum value for slow-start threshold
(ssthresh) allows connection bandwidth utilization to ramp
up as aggressively as possible without overshoot.
No reliable way to measure ssthresh!

25
cwnd (segments)

20
15 ssthresh

10
5
0
0 1 2 3 4 5 6 7 8
RTT
 Delayed Duplicate ACKs

The delayed duplicate acknowledgement scheme

selectively delays duplicate acknowledgments at the
receiver.

Not so beneficial when link layer guarantees in

sequence delivery (link layer residual errors are very
small).
 Delayed Duplicate ACKs
K
K+1
K+1
K+2
K+2
K+3
K+4
K+5
K+6 Delay
K+7
Fast RTX K+2 K+2
K+8 K+8
Fast Recovery K+9 K+9
K+10
 Selective Acknowledgments (SACK)

Limited information available from cumulative

acknowledgments.TCP sender can only learn about a single
lost packet per round trip time

With SACK, the receiving TCP sends back SACK packets to

the sender informing the sender of data that has been
received. The sender can then retransmit only the missing
data segments

Useful in cases with considerable probability of multiple

segment losses per window.
Selective Acknowledgments
K
K+1
K
K+2
K+1
K+3
K+4
K+5
K+6
K+7 (K+3, K+4, K+6)
K+8
K+2
K+5
K+9
 Split TCP

Replace an end-to-end TCP connection with two distinct

connections: one across the wireless link, the other across
the wireline network

TCP1 TCP2

Host BSC/RNC BS MS
 Split TCP: Advantages

Each of the connections can be optimized

independently.

For example, packet losses due to congestion and

transmission error can be distinguished.

Allows earlier deployment of enhancements to TCP

over wireless.

Allows exploitation of various application level

enhancements.
 Split TCP: Drawbacks

Breaks TCP end-to-end semantics e.g., disables end-to-

end usage of IP layer security mechanisms.

Crashes of the intermediate node (containing TCP state

machine in the radio access network) become
unrecoverable.

Extra overhead in moving TCP states in case of handoff

between the intermediate nodes.

Extra processing overhead at the intermediate node.

 Snoop Protocol

Locally (on the wireless link) retransmit lost packets.

Suppress the duplicate ACKs on their way from the
receiver back to the sender.
Not beneficial when radio link protocol provides in
sequence delivery of frames.
Problem with IPSEC.

Wireless Link

Internet

Host BSC/RNC BS MS
 Application Level Proxies

Employing an application level proxy practically

results in a split TCP connection.

The proxy function as the intermediary.

Avoids overhead on the wireless link e.g., round trips,

headers and inefficient coding associated with
application protocols.

Examples: Web proxy caches, relay MTAs (Mail

Transfer Agents) and SOCKS etc.
 Performance Enhancing Proxy

Wireless links suffer from periods of disconnection due to

handoffs and unavailability of resources etc.

TCP retransmits data on retransmission timeout (RTO).

This causes TCP to close its congestion window.

 Performance Enhancing Proxy

Introduce a supervisor host (SH-TCP) at the edge of

the wireless link (e.g., at GGSN in GPRS).

Retain the last ACK coming from the receiver.

During periods of disconnection, the SH-TCP can

shut down the senders window by sending the last
ACK with a window set to zero.

This make the sender go to persist mode.

 HTTP/TCP/RLP Results

ns (network simulator) tool.

Web browsing traffic model [Bruce Mah].
HTTP/1.0.
TCP Reno.
IP/PPP overhead taken into account.
IS-2000 forward and reverse link data rates.
One-way Internet delay is fixed at 80ms.
Correlated and i.i.d radio link errors.
 HTTP Traffic Model

Web Browsing Session

Web Page Web Page Web Page

Think Time
Mouse click Mouse click Web objects
 HTTP Traffic Model Parameters

HTTP Request Size One large peak around

250 bytes and a small
peak around 1KB
Number of Mean = 3,
connections per web Median=1
page
Web object size Mean = 6.75KB

Think Time between 30* seconds

pages [Exponential distribution]
 TCP Parameters

TCP version Reno

TCP maximum 576 or 1500 bytes

segment size (MSS)

TCP receiver window 32KB

size
 Radio Link Parameters

RLP NAK scheme Configurable

One-way radio link 100ms

delay

RLP guard time 60ms

Radio link data rates 9.6, 19.2, 38.4, 76.8 and

153.6 Kb/s.
 HTTP Performance Measures

Web Object Transfer Time [secs]

Normalized Web Object Delay [secs/KB]

Web Page Transfer Time [secs]

Normalized Web Page Delay [secs/KB]

 Web Object Transfer Time

Object Size: 1-10KB, Data Rate = 38.4 Kb/s

5
4
Delay [secs]

3 C=1
2 C=4
1
0

r)
r)

r)
)

)
I.d
I.d

I.d

or

or
or
(I.

(I.

(I.

(c

(c

(c
01

05

10

01

05

10
0.

0.

0.

0.

0.

0.
FER

C: Maximum number of parallel TCP connections in HTTP/1.0

 Normalized Object Delay
Radio link data rate = 38.4 Kb/s, C=1

4
3.5
Delay [s ec s / K by te]

3
2.5 0.01(I.I.d)
2 0.05(I.I.d)
1.5 0.10(I.I.d)
1
0.5
0
< 1K B 1K -10K 10K - 20K - 30K -
20K 30K 40K

O bjec t S iz e
 O bjec t Throughput [K b/s ]
0.

10
20
30
40
50
60

0
01
( I.
I. d
0. )
05
(I.
I. d
0. )
10
(I.
I. d
)
0.
01
(c
or

FER
r)
0.
05
(c
or
r)
0.
10
(c
Object Size: 20-30KB, C=1
Object Throughput

or
r)
38.4 K b/s
153.6 K b/s
 P age Throughput [K b/s ]
0.

20
40
60
80
100
120
140

0
01
(I.
I. d
)
0.
05
( I.
I. d
)
0.
10
(I.
I. d
)
0.
01
(c
or

FER
r)
0.
05
(c
or
r)
0.
10
(c
Radio link data rate = 153.6 Kb/s
Web Page Throughput

or
r)
C= 4
C= 1
 Normalized Web Page Delay
Radio link data rate =153.6 Kb/s

8
7
Delay [s ec s / K by tes ]

6
5 0.01(c orr)
4 0.05(c orr)
3 0.10(c orr)
2
1
0
< 1K B 1K -10K 10-100 100-500 500-1M

O bjec t S iz e
 HTTP Performance Summary

Allowing multiple simultaneous TCP connections for a

web page transfer reduce the page transfer times while
increasing the individual object transfer times.

Increasing the air link bandwidth will reduce the transfer

times for large objects/pages (>1KB) only.

For smaller objects the transfer delay is lower bounded

by HTTP request/response delay involving multiple
RTTs.
 System Capacity with HTTP Traffic

180 1

Normalized Delay [secs/KB]

Sector Throughput [Kb/s]

160 0.95

140 120 Kb/s 0.9

0.82 s/KB
0.85
120
0.8
100
0.75
80
0.7
60
0.65
40 0.6
20 0.55 26 users
0 0.5
0 10 20 30 40 50 0 10 20 30 40 50

Number of HTTP users Number of HTTP users

 RLP NAK Schemes

The Schemes vary in the total number of

retransmission rounds and the number of rounds in
each round.

Notation (n1, n2,, nk) denotes k rounds of

retransmissions with n1, n2 and nk NAKs in round 1,2
and k respectively.

The total number of retransmission attempts is

limited to n1+n2 +,.., +nk
Effect of FER on HTTP Performance

Normalized Delay [secs/KB] OS=10-20KB, i.i.d. errors

1.4
1.2 (1,1,1,1,1)
1 (2,3)
0.8 (1,1,2)
0.6 (1,1,3)
0.4 (1,2,2)
0.2 (1,2,3)

0
%
0%

%
=5

10

20

30
R=

R=

R=

R=
R
FE

FE

FE

FE

FE
Effect of FER on HTTP Performance

Normalized Delay [secs/KB] OS=10-20KB, Correlated errors

1.4
1.2 (1,1,1,1,1)
1 (2,3)
0.8 (1,1,2)
0.6 (1,1,3)
0.4 (1,2,2)
0.2 (1,2,3)
0

%
0%

5%

30
10

20
R=

R=

R=

R=
R=
FE

FE

FE

FE

FE
 RLP Buffers

TCP Segments

TCP Segments New Transmission

buffer

V(R) -V(N)

RESEQ Buffer
V(S) -V(N)peer

RTX buffer

MS BSC/RNC
RLP new TX Buffer Size C.D.F.

1
0.9
0.8
0.7
0.6
C.D.F.

FER=0%
0.5
FER=10%
0.4
0.3
0.2
0.1
0
0
0.86
1.73
2.59
3.46
4.32
5.18
6.05
6.91
7.78
Buffer Size [KB]
 Over-the-air Packet Multiplexing

With a single TCP connection for an HTTP page transfer,

the RLP new transmission buffer is empty for 58% and
63% of the time for the case of no errors and FER of 10%
respectively.

This explains the need for packet multiplexing over the air.

A circuit switched dedicated channel will make inefficient

use of the scarce radio bandwidth.
 TCP over W-CDMA

draft-inamura-docomo-00.txt

Choice of MTU size in link layer

Appropriate receive window size

Increased initial congestion window

Use of Selective ACK

Effect of TCP Receive Window

Throughput [Kb/s]

BLER 32KB 64KB 128KB

1% 340 369 369

5% 275 340 340

10% 230 320 320

 TCP over W-CDMA
Feature Recomm Comments
endation

MTU size 1500B Larger MTU helps TCP

window grow faster.
Small MTU is better for
high BLERs.
TCP Receive 64KB Going from 64KB to
Window size 128KB does not
improve performance
TCP Initial 2MSS Effective for small data
Window (up to transfers
4380B)
Use of SACK Yes Useful when multiple
TCP segments are lost
in a single window
 Summary

Packet Multiplexing over the air is crucial for efficient support

of TCP/IP traffic.

A radio link layer MUST perform in sequence delivery of radio

frames to avoid TCP retransmissions due to duplicate ACKs.

Multiple instances of link layer protocol per user (e.g., for

different TCP flows) will improve the overall system
performance (assuming low overhead to setup a radio link
connection)
 Summary (Continued)

Minimizing the radio link round trip time (RTT) by

eliminating the delays associated with the access network,
coding and interleaving etc. will improve the performance
of TCP/IP traffic.

Limit the radio link errors to =< 10%

Use less aggressive NAK schemes.

 Thank you !!!

For more information, please send e-mail to

Farooq Khan at
fkhan1@lucent.com