QoS IMPROVEMENTS RESULT FROM TCP/RLC AND MAC IN

A MOBILE CHANNEL

Jahangir Dadkhah Chimeh, Mohammad Hakkak

Islamic Azad University Science and Research Branch, Hesarak, Tehran, Iran

Tarbiat Modares University, Jalale-Aleahmad, Tehran, Iran

Hamidreza Bakhshi, Paeiz Azmi

Shahed University,Khalij Fars highway, Tehran, Iran

Tarbiat Modares University, Jalale-Aleahmad, Tehran, Iran

Keywords: TCP/ARQ, UMTS, fading channel.

Abstract: Mobile telecommunication new services are based on data networks specially Internet. These services

include http, telnet, ftp, Simple Mail Transfer Protocol (SMTP), etc. Besides we recognize a mobile network

as a multi-user network. Transmission Control Protocol/Internet Protocol (TCP/IP) which is sensitive to link

congestion in wireline data links is also used in wireless networks. In order to improve the system

performance, the TCP layer uses flow control and congestion control. Besides, Radio Link Control (RLC)

has been introduced to compensate the deficiency of TCP layer in wireless environment. MAC and RLC

have important roles in quality of service improvement of UMTS. In this paper we verify TCP over

Automatic Repeat reQuest (ARQ) error control mechanism and finally quality of service improvement

results from it in the fading channels.

1 INTRODUCTION

Transmission Control Protocol (TCP) is an end-to-

end transport protocol in the Internet Protocol (IP)

suite which is widely used in popular applications

like SMTP, ftp and http. TCP guarantees reliable

and in-sequence delivery of packets. TCP

performance gets severely affected when used on

channels which are typically characterized by high

error rates (e.g., wireless channels). Although Tahoe

and Reno versions of TCP are commonly used

implement-tations, several other enhancements to

TCP, including New Reno, Selective

Acknowledgement (SACK), and Vegas, have been

proposed to improve TCP performance. However, it

has been shown that TCP enhancements like New

Reno and Vegas do not offer significant

improvement compared to Tahoe on slowly fading,

low bandwidth-delay product wireless channels. It

was suggested by (Chockalingam and Zorzi, 1999)

that a clever design of the lower layers where

memory in the packet error process (naturally

present on wireless links because of the fading

behavior) is preserved could be more effective than

the development/use of more complex TCP versions.

Wireless radio channel is affected by shadowing and

fading. These, in turn, are affected by environment

and mobility of subscribers. On the other hand,

developing data networks and their connections to

the mobile network has introduced new attractive

services. These services are in addition to customary

voice services. Data services include data file, video

and voice transfer within the mobile network or over

the Internet connection. Internet traffic is a

combination of the above traffic services.

We know data traffic is sensitive to packet loss

and voice traffic is sensitive to delay

(Chockalingam, 2000). TCP provides end-to-end

recovery and flow control. Link layer recovery

protocols in cellular systems operate below the TCP

layer and have complex interactions with TCP.

TCP has been designed and tuned for networks

in which segment losses and corruptions are mainly

due to congestion. TCP utilizes a flow control

mechanism by which the sender host doesn’t

overflow the receiver host by transmitting too much

Dadkhah Chimeh J., Hakkak M., Bakhshi H. and Azmi P. (2008).

QoS IMPROVEMENTS RESULT FROM TCP/RLC AND MAC IN A MOBILE CHANNEL.

In Proceedings of the International Conference on Signal Processing and Multimedia Applications, pages 31-34

DOI: 10.5220/0001931400310034

 SciTePress

data and too fast. In order to avoid long delays when

there is no response from the receiver in a TCP

connection, a time-out mechanism is employed.

Besides a congestion control mechanism is used in

TCP to avoid packet drop due to lack of resources

and buffer space.

In wireless systems most of errors are due to lossy

media. The reason is that in the wireless channels

the main cause for the packet loss may be the high

BER in the channel not the network congestion. So

the low efficiency of the TCP in a wireless channel

is a direct result of the fact that the TCP

misinterprets the packet loss resulting from high

channel error rate or from the congestion. In order to

enhance QoS seen by TCP layer on a wireless link, a

radio link control (RLC) is generally introduced at

link layer. Typically the RLC uses an ARQ error

recovery mechanism to improve the QoS (3GPP TS

25 322, 2007).

RLC is a protocol above MAC and blew RRC.

Every outgoing TCP packet is put into an interface

buffer which is picked up by the RLC. RLC is

responsible for error and flow control (by ARQ

mechanism) of the frames and provides transparent

mode (TM), unacknowledged mode (UM) and

acknowledged mode (AM) services. The RLC

breaks the TCP packet into 10-ms frames and sends

them to MAC. MAC chooses a user queue according

to the scheduling mechanism and after adding a

MAC header sends them to the Physical layer

(Chockalingam and Zorzi, 1999) and (Borgonovo,

2001). In this paper we review the ARQ protocol

effects on the TCP throughput and show how it

improves the throughput. In section 2 we define a

system model and in section 3 we simulate the TCP

and TCP/ARQ protocols in wired and wireless

systems. Finally conclusion will be offered.

2 SYSTEM MODEL

TCP is in layer 4 and locates in the hosts in end

nodes but isn’t a part of UMTS network.

Implementations of TCP contain four intertwined

algorithms: slow start, congestion avoidance, fast

retransmit and fast recovery (RFC, 2001). Although

TCP has been designed, optimized and tuned in

wired networks to react to the packet loss due to

congestion, in wireless systems service degradation

can be due to bit (packet) errors. In UMTS, TCP and

ARQ protocols operate against loss and error in

wired and wireless sections respectively.

TCP in a wireless network experiences several

challenges. One of the issues is how to deal with the

spurious timeout caused by the abruptly increased

delay, which triggers unnecessary retransmission

and congestion control. It is known that the link-

layer error recovery scheme, the channel scheduling

algorithm, and handover often make the link latency

very high. Bandwidth of the wireless link often

fluctuates because the wireless channel scheduler

assigns a channel for a limited time to a user. Thus,

the variance of inter-packet arrival time becomes

high, which may result in spurious timeout. The

Eifel algorithm has been proposed to detect the

spurious timeout and to recover by restoring the

connection state saved before the timeout

(Wennstrom, 2004) and (Gurtov, 2003).

Although the packet loss rate of the wireless link

has been reduced due to link-layer retransmission

and Forward Error Correction (FEC), losses still

exist because of the poor radio conditions and

mobility. Therefore, non-congestion errors could

sharply decrease the TCP sending rate. Packet

reordering at the TCP layer may be caused by link-

layer retransmission, which also results in

unnecessary retransmission and congestion. In the

wireless networks, in general, bandwidth and latency

at uplink and at downlink directions are different.

Hence, the throughput over downlink may be

decreased because of ACK congestion at the uplink

(Lee, 2006).

Now we consider a TCP connection between two

hosts such that the first link on the end-to-end path

from the sender to the receiver is a wireless radio

link (Lee, 2006) and (Canton 2001). Such a scenario

is common in mobile communication and is

illustrated in figure 1(a). The protocol stack on the

way from mobile host to fixed host is illustrated in

the figure 1(b).

We assume there is no packet loss due to

congestion on the wireless link but some packets

may be corrupted under adverse radio link

conditions. In our study, we assume that the bit error

patterns on the radio link are independent. On the

wired network, packets may only get lost when

congestion occurs.

As described in (Lee, 2006) and (Chahad, 2003)

we assume that TCP sends one cumulative

TCP

ACK for b consecutive TCP segments and is

always in congestion avoidance. Besides, Packet

loss is detected in one of the two ways, either upon

reception of a triple-duplicate

TCP

ACK (denoted

by TD), or upon expiration of a Time-Out (denoted

by T0). In case of a TD, window size is decreased by

half, while upon expiration of a T0, it is decreased to

1. Moreover, we assume that the loss behavior is

SIGMAP 2008 - International Conference on Signal Processing and Multimedia Applications

bursty, i.e., packet losses are correlated within a

back-to-back transmission. Hence, when a packet is

lost, all remaining packets in the same round are lost

as well (Padhey, 1998). Furthermore, under the

assumption that rounds are separated by each TCP

round trip time,

TCP

TT , loss in one round is

independent of loss in other rounds.

(a)

(b)

Figure 1: (a) An end-to-end system (b) The protocol stack

on the way.

We consider TCP Reno version. Let T0 denote the

TCP time-out and p denote the loss rate due to

congestion in the wired portion of the network. For

the steady-state of TCP throughput, in a wired

context only we have (Canton, 2001) and (Chahad,

2003)

()

min(1,3 ) (1 32 )

TCP

Th p

bp bp

TT T p p

(1)

and for an end to end protocol consisted of both

wireline and wireless channel we have

()

⎟

⎠

⎞

⎜

⎝

⎛

−++

×−+

−+

−

)*(321

)*(3

3,1

min

)*(2

)],([

PERpPERp

PERpPERpb

RTTPERpTh

TCP

(2)

which is the same as (1) in which p is substituted by

1(1 )(1 ) *

Global Average Packet Loss Rate

PER p p PER p PER

−− − =+ −

(3)

In addition we have packet error rate as in

equation

1(1 )

PER FER=− −

in which FER is frame

error rate in wireless section and n is the number of

frames in a packet. Equation (2) is for a complete

(wireline and wireless) system without ARQ

mechanism.

If we consider a Go Back-N mechanism in layer 2

and also the case of independent and identically

distributed bit errors (i.i.d.), we have the throughput

as (Wennstrom, 2004)

)321()

3,1(

min

)1(

)],([

FER

nbFER

RTTnbDRTT

FERpTh

ARQ

wlessARQwire

⎟

⎠

⎞

⎜

⎝

⎛

−

+++

−

(4)

in which

ARQ

D is the constant delay component as a

result of ARQ frame processing and

wless

RTT and

wire

TT are round trip times for ARQ (air channel

from UE to RNC and vice versa) and wired sections

respectively.

Now we consider a mobile channel where a

subscriber moves in it. It is surrounded by some

obstacles which incident rays strike them. We

compute BER as described in (Jeruchim, 2000)

which has been used in MATLAB as a reference for

fading channel simulation. There, it is modeled as a

linear FIR filter with tap weights given by

)(sin NnNfornTchg

kkn

≤≤−−=

∑

(5)

where

- the summation has one term for each major path.

- {

} is the set of path delay.

- T is the input sample period.

N and

N are chosen so that

g is small when n

is less than -

or greater than

- {

h } is the set of complex path gain which are not

correlated with each other.

3 SIMULATION

We used RLC in acknowledged mode (AM) with a

Go-Back-N mechanism. Besides we used a TCP

packet with 120 bytes length then segmented it into

10 frames with 12 bytes length each. We began

finding the ordinary TCP throughput in the wired

section. The TCP parameters are

4.00102.0

TbRTT

wire

. Then by MATLAB

we made a slow fading channel and calculated the

QoS IMPROVEMENTS RESULT FROM TCP/RLC AND MAC IN A MOBILE CHANNEL

Figure: 2 Effect of air channel on TCP throughput (kbps).

Figure: 3 Effect of air channel on TCP/ARQ throughput

(kbps).

TCP throughput in it. Here we considered a

pedestrian user with the speed of 3m/s (slow fading,

Hzf

20=

), then we considered a mobile user with

the speed of 81km/h (fast fading,

Hzf

100=

) and

found the throughput as shown in Figure 2. It

consists of three kinds of curves. These are related to

wired, slow and fast fading channels with No ARQ

protocol. We see fading and FER effects on the

throughput. The least throughput is related to a fast

fading multipath channel where a user moves with

the speed of 81km/h.

Then in another scenario like mentioned above

we found the TCP/ARQ throughput with Go-Back-

N mechanism as shown in Figure 3. We used Go-

Back-N mechanism with the following parameters.

0.005 0.001 10 10

ARQ ARQ

RTT D n N====

ARQ protocol effects on throughput in slow and fast

fading channels are illustrated in figure 3. The

throughput improvement related to ARQ protocol is

obviously observed on it.

4 CONCLUSIONS

We could compensate the TCP deficiency in

wireless channels by ARQ protocol in the layer two.

Besides we observed that although a fast fading

channel degrades the system throughput more than a

slow fading channel but ARQ protocol improves the

throughput more.

REFERENCES

Chockalingam, A., Zorzi, M., 1999. V. Tralli, Wireless

TCP Performance with Link Layer FEC/ARQ, IEEE

International Conference on Communications,

ICC'99, Volume 2, 6-10 Page(s):1212 – 1216.

Chockalingam, A., 2000. Performance of TCP/RLP

Protocol Stack on Correlated Fading DS-CDMA

Wireless Links, IEEE Transactions on Vehicular

Technology, Vol. 49, No 1.

3GPP, TS 125 322, 2007. Radio Link Control (RLC)

protocol specification.

Borgonovo, F., Capone, A., Cesana, M., Fratta, L., 2001.

Delay-Throughput of Packet Service in UMTS, IEEE

VTS 54

, Vol4, Page(s): 2128-2132.

RFC 2001, TCP Slow Start, Congestion avoidance, Fast

retransmit and Fast recovery algorithms.

Wennstrom, A., Alfredsson, S., Brunstorm, A., 2004. TCP

over Wireless networks, Karlstad University Press,

Sweden.

Gurtov, A., Ludwig, R., 2003. Responding to Spurious

Timeouts in TCP, IEEE INFOCOM.

Lee, Y., 2006. Measured TCP Performance in CDMA 1x

EV-DO Network, PAM conference, Adelaide,

Autralia.

Canton, A.F, Chahed, T., 2001. End-to-End Reliability in

UMTS: TCP over ARQ, Globecom’2001, San

Antonio.

Chahad, T., Canton A. F., Elayoubi, S. E., 2003. End-to-

end TCP Performance in W-CDMA / UMTS, ICC’03,

IEEE, Vol1, Page(s): 71-75.

Padhey, J., Firoiu, V., Towsley, D., Kurse, J., 1998.

Modeling TCP Throughput: A Simple Model and its

Empirical Validation, Proc. ACM SIGCOMM’98,

pp.303-314.

Jeruchim, M.C., Balaban, P., Shanmugan, K. Sam, 2000.

Simulation of Communication Systems: modeling,

methodology and techniques, Kluwer Academic /

Plenum Publishers.

SIGMAP 2008 - International Conference on Signal Processing and Multimedia Applications