Download Buffer Allocation in Wireless Multimedia Networks Using

Home | Sessions | Authors | Session 6.2
Buffer Allocation in Wireless Multimedia Networks Using Fuzzy Logic
J. D. Mallapur, S. S. Manvi, R. B. Manjula
D. H. Rao
Basaveshwar Engineering College
Bagalkot, Karnataka, INDIA
Gogte Institute of Technology
Belgaum, Karnataka, INDIA
[email protected],
[email protected]
Abstract
In future generation cellular systems, network resources
like bandwidth, buffers, computing power, etc., have to
be allocated with an efficient resource allocation method
because of their scarcity. This paper presents a fuzzy
based buffer allocation scheme for wireless multimedia
networks in the context of future generation cellular networks. In this scheme, a buffer allocator located at the
base station uses a fuzzy controller for buffer allocation
that uses three fuzzy parameters namely application priority (based on handed and new calls), transmission rate,
and packet size. Fuzzy controller computes an allocation
factor, which is used to allocate the buffers to an requesting application. The scheme has been simulated in different network scenarios to test the operation effectiveness.
The simulation results show that fuzzy based buffer allocation scheme performs better than conventional scheme
in terms of buffer utilization, handoff and new calls acceptance.
1. Introduction
Wireless multimedia networks are the need of today’s society. There is tremendous demand for multimedia applications over wireless cellular networks. Some of the
applications are, video conferencing, Internet multimedia games, and e-commerce. To provide all such applications in wireless networks, we need to look at some of the
pressing issues which arise due to scarcity of resources
such as bandwidth, buffers, battery life, etc.
Wireless and mobile links supporting packet data
transfer require buffers to accommodate inherent load
variations of various traffic and insufficient bandwidth
allocated to a user. Issues in buffer allocation can be
stated as follows. On one side, such a buffer needs to
be sufficiently large to ensure good link utilization and
low blocking probability, on the other side it needs to be
small to minimize queuing delays. Hence there should be
some buffer allocation technique which can take care of
reducing delays as well utilize the links efficiently and accept more number of users in a cell. This paper presents a
fuzzy based buffer allocation scheme by considering the
fuzzy parameters such as priority of the application, rate
[email protected]
of transmission and packet size.
Some of the works on buffer allocation for wireless
multimedia networks are as follows. In [4], adaptive filter based dynamic buffer allocation method is presented
that provides better Quality of Service (QoS) to an requesting application. A technique to reduce the waiting
time portion of query processing through flexible buffer
allocation is presented in [3]. An evaluation of a active
queue management scheme tailored to specific characteristics of 3G links is carried out in [6]. A new queue
management algorithm in order to utilize high data rate
efficiently in current wireless LAN is given in [5]. In [7],
resource allocation is done by using fuzzy logic, where
network preferentially allocates its resources to real-time
(RT) traffic sources and optimally allocates the remaining
resources to NRT traffic based on channel fading conditions. An active queue management to limit the transmission queue length that eliminates expiration packet drops
is discussed in [1]. A random early expiration detection
based on buffer management algorithm for real-time traffic over wireless networks is presented in [2].
The rest of the paper is organized as follows. Section 2 describes the proposed fuzzy based buffer allocation scheme. Section 3 presents simulation model used to
analyze the fuzzy based buffer allocation scheme. Results
are presented in Section 4. Finally, Section 5 concludes
the paper with some remarks.
2. Proposed Work
The potential of mobile multimedia services is enormous. Another major challenge is the characterization
of the resource (bandwidth, buffer, consumption power
etc) requirements of multimedia documents. The multimedia communication application requires a class of network resources like high bandwidth, buffer and computing power. If the bandwidth required by each application
is not available, then buffer plays a vital role. Hence for
efficient utilization of the buffers placed at the base station, we propose a fuzzy based buffer allocation.
In this section, we present the network environment,
fuzzy based buffer allocation scheme and the algorithm
for buffer allocation.
2.1. Network environment
We consider a cell with several users operating in it.
Users may come from other cells adjacent to a cell thus
creating handoff calls. New calls can be generated within
a cell with certain application requirements. Buffer allocation scheme exists at the base station, which allocates buffers to handoff/new calls based on certain criteria. Buffers at the base station are divided into two
categories: handoff call and new call buffers. For handoff calls, network will first make an attempt to allocate
handoff calls within handoff buffers, if not available, it
will try to allocate from new call buffers. Decision to fix
handoff call buffers is decided by the cell administrator
based on history of call arriving into a cell from other
cells. Some of the users may move out of the cell thus
releasing buffers, which can be used for other requesting
applications.
APPLICATION
REQUEST
BUFFER
ACCEPT/REJECT
ALLOCATOR
APPLICATION
DATABASE
FUZZY
ALLOCATION
Figure 1: Buffer allocation model
2.3. Fuzzy controlled buffer allocation
Fuzzy controlled buffer allocation is shown in figure 2,
which consists of fuzzification, inference, and defuzzification steps.
PRIORITY
PACKET SIZE
FUZZIFIER
INFERENCE
RATE OF
BUFFER
DEFUZZIFIER
ALLOCATION
FACTOR
FLOW
2.2. Buffer allocation
The proposed fuzzy based buffer allocation is shown in
figure 1 which is located at the base station, which comprises of buffer allocator, application database and fuzzy
allocation scheme. The functions of each block is given
below.
• Buffer Allocator: it receives application connection request either from handoff/new calls with required specifications such as bandwidth, delays,
etc. Based on the type of call it first tries to allocate from relevant buffers (hand off/new) or any of
the free buffers if required bandwidth is not available by using fuzzy allocation scheme. If sufficient buffers are allocated to accommodate at least
minimum bandwidth requirement of an application, then application will be accepted, otherwise
rejected. It updates the application database with
call information and the buffers allocated to it.
• Fuzzy allocation: this calculates the allocation factor for requesting application and sends the calculated allocation factor to the buffer allocator to efficiently allocate the buffer. Care is taken that handoff and real-time calls are given highest preference
than any other type of call.
• Application database: this comprises of information of all existing calls such as bandwidth
required, bandwidth allocated, buffers allocated,
buffer allocation factor, type of call, etc. It also
comprises of cell status such as bandwidth available, buffers available in both handoff and new call
categories.
RULE BASE
Figure 2: Fuzzy buffer allocation factor computation
In the fuzzification step, fuzzy parameter values are
converted into linguistic values (such as low, high or
medium). Each fuzzy set is associated with a membership function used to characterize how certain the crisp
input belongs to the set. For a given crisp input, the membership function returns a real number in the range [0,1].
The closer the membership value is to 1, the more certain the input belongs to the set. Fuzzy inputs considered
in the proposed work are priority, packet size and rate of
flow. A single crisp value can take more than one linguistic value if the membership values overlap. In the
inference step, a set of rules called rule-base, which emulates the decision-making process of a human expert is
applied to the linguistic values of the inputs to infer the
output sets which represents the actual control signal for
the process. We refer the reader to [9] for more complete
background information on the fuzzy control.
2.4. Fuzzification
Fuzzy based buffer allocation scheme considers three parameters for fuzzification: priority of the application (P),
packet size of each application (PS) and the rate of flow
of each application (R). The output of linguistic parameters is the buffer allocation factor for a given application.
The membership to each of fuzzy variables is assigned
using intuition method. For each of the considered fuzzy
parameter, their range of linguistic values are depicted in
figure 3.
LOW
MEDIUM
HIGH
priority
priority
P0
P1
SMALL
P2
P3
MEDIUM
P4
LARGE
packet size
ps0
ps1
LOW
ps2
ps3
ps4
HIGH
MEDIUM
rate flow
HG
HI
HG
HG
HG
HG
HG
HI
HI
MD
MD
MD
HG
LW
HG
HG
ME
LW
LW
HI
ME
HI
ME
r1
r2
r3
MEDIUM
LOW
r4
HIGH
buffer allocation
factor
bf0
bf1
bf2
bf3
bf4
Figure 3: Membership function for input and output linguistic parameters
2.5. Inference and defuzzification
Since there are three linguistic values P , P S and R, the
total number of rules is 27. If the condition is true, we
call the rule as being active. In our case, the rule-base is
in a form called functional fuzzy system where each rule
i is written as follows.
Rule i:IF P is low and R is low and P S is
small,THEN AF = low
Where AF linguistic value is decided based on membership functions of three input fuzzy parameters, priority, packet size and flow rate. To decide an appropriate
AF , the strength of each rule must be considered. For
this reason, the output membership function is a complicated function and center of area method [9] is used for
defuzzification. This method finds the center point of the
fuzzy output membership function which is used for allocating buffer for requesting application. The fuzzy rule
base with 27 rules is shown in figure 4.
The defuzzified output parameter will give flexibility
to the network administrator to perform soft buffer allocation.
Packet size
Allocation factor
LR
MI
H
S
LR
MI
S
LR
MI
S
H
M
M
M
M
M
M
LR
M
MI
M
HI
S
M
ME
MD
LR
M
ME
MD
MI
M
ME
ME
ME
MD
LW
S
M
ME
LW
HI
HI
LR
MI
S
L
L
L
M
M
LO
LO
LO
r0
Rate of flow
LW
H
HI
LR
MI
S
LO
MD
LR
M
L
LO
MD
MI
L
LO
S
L
LO
LO
MD
LW
LW
LR
MI
L
L
LO
LW
S
L
PRIORITY (HG=HIGH,ME=MEDIUM,LO=LOW)
RATE OF FLOW(LW=LOW,MD=MEDIUM,HI=HIGH)
PACKET SIZE (SMALL=S,MI=MEDIUM,LR=LARGE)
ALLOCATION FACTOR(L=LOW,M=MEDIUM,H=LARGE)
Figure 4: Fuzzy rule base table
buffers reserved for handoff calls and y% for new calls
generated within the cell.
Algorithm 1: Buffer allocation in a base station
{Nomenclature: n= number of requesting applications, BW max= Maximum bandwidth,BW req= Bandwidth requested (in Mbps), Bav= Available bandwidth, Bmax=Maximum buffer size at the base station,
N ormal Buf f =y*Bmax, Handof f Buf f =x*Bmax,
i=ith running application, t = time for buffering, Breq=
Buffer required in one second, BALLOC = Buffers allocated, AF = Allocation factor, SU M = Buffers allocated
for existing calls within the base station, SU M 1= Buffers
allocated for handoff calls.}
Begin
1. Receive the application request with required bandwidth, BWreq;
2. If (BWreq ≤ Bav) then bandwidth is allocated,
Else compute the buffer required, which is equal
to, Breq = (BWreq - Bav)*t;
3. Call Algorithm 2 to compute buffer allocation factor (AF);
4. For new calls perform the following
2.6. Algorithm
• BALLOC= Breq * AF;
This section presents pseuocode (Algorithms 1 and 2) for
the working of the proposed scheme. We consider x%
• If BALLOC ≥ (Normal buff - SUM) then reject the new call;
• If BALLOC ≤ (Normal buff - SUM) then allocate the buffers and update the application
database with allocated buffers and increment
SUM by BALLOC; inform the application;
5. For handoff calls perform the following
• BALLOC= Breq * AF;
• If BALLOC ≥ (Handoff buff - SUM1) then
allocate from Normal buff if buffers available, otherwise, reject the handoff call;
• If BALLOC ≤ (Handoff buff - SUM1) then
allocate the buffers and update the application
database with allocated buffers and increment
SUM1 by BALLOC; inform the application;
6. Stop
End.
Algorithm 2: Computation of allocation factor
Begin
BWreq1=3, BWreq2=5, pb=0.2, Bmax=100 and 200
for different cases, Bn=80%, Bh=20%, r0=1.0, r1=1.25,
r2=1.5, r3=1.75, r4=2.0, ps0=500, ps1=750, ps2=1000,
ps3=1250, ps4=1500, bf0=0.25, bf1=0.45, bf2=0.65,
bf3=0.80, bf4=1.0. Simulation procedure is as follows.
Begin
1. Generate a cellular network.
2. Generate the application/call requests.
3. Apply the proposed scheme.
4. Compute the performance of the system.
End
The performance parameters measured are as follows.
• Buffer utilization: It is defined as the ratio of buffer
utilized to the maximum size of buffer available at
base station.
1. Initialize fuzzy controller with priority of a application, rate of flow of data, packet size of a application;
• Handoff Calls Accepted: It is defined as the ratio
of handoff calls accepted to the total handoff calls
arrived.
2. Find the membership function of priority, rate of
flow and packet size allocated to application;
• New Calls Accepted: It is defined as the ratio of
new calls accepted to the total new calls generated.
4. Inform AF value to buffer allocator;
5. Return;
6. Stop;
End.
3. Simulation
This section describes the simulation model, simulation
procedure and performance parameters. Simulation is being carried out on a pentium-4 machine by using C programming language.
A single cell environment with an area of (x, y) meters is considered. n number of users are generated in
a cell comprising of both handoff and new calls. Maximum bandwidth of a cell is assumed to be BW max
Mbps. Bandwidth requests for each call are generated
randomly in the range BW req1 to Bred2 Mbps. Calls
are randomly categorized into handoff and new calls by
using a probability pb, i.e., if generated random number
(0 to 1) is less than pb, then it is called as handoff call.
Maximum Buffer Size at the base station of the cell is assumed to be Bmax Mbytes . Maximum Buffer size is divided into two regions, Normal Region= Bn%of Bmax
MBytes and Handoff Region= Bh%of Bmax MBytes.
Following inputs are considered for simulation.
x=500, y=1000, n is varied from 5 to 50, BWmax=20,
3.1. Results
The results that we will present here demonstrate the capability of our fuzzy buffer allocation algorithm to allocate the buffer in worst case. Also, we compare the fuzzy
based buffer allocation scheme with non fuzzy based
buffer allocation scheme. A non fuzzy based buffer allocation scheme performs static buffer allocation based
on the absolute buffer requirements besides considering
random early expiration detection for packet dropping.
From figure 5, we observe that buffer utilization for different capacity of buffers (100 and 200 MB) as compared
to non fuzzy method is better. This is because of soft
buffer allocation based on application type.
% of buffer utilization .Vs. Number of calls
1
0.9
% of buffer utilization
3. Compute AF by referring to rule base;
0.8
0.7
0.6
0.5
Fuzzy (100 MB)
0.4
Non fuzzy (100MB)
Fuzzy (200 MB)
Non fuzzy (200MB)
0.3
0.2
0.1
5
10
15
20
25
30
35
40
Number of calls
Figure 5: Buffer utilization (%) Vs. Call arrival rate
It is seen that fuzzy based scheme has better acceptance of handoff calls for both the cases of buffer capacity
as depicted in figure 6.
5. Acknowledgment
% of accepted handoff calls .Vs. Number of handoff call arrivals
We are thankful to All India Council of Technical Education, New Delhi, and National Project Implementation Unit, New Delhi, for sponsoring the
part of the project under Research Promotion Scheme
(8023/RID/BOR/RPS-43/2005-06) and TEQIP, respectively.
% of accepted handoff calls
1
0.9
0.8
0.7
0.6
Fuzzy (100 MB)
Non fuzzy (100MB)
Fuzzy (200 MB)
Non fuzzy (200MB)
0.5
0.4
0.3
2
4
6
6. References
8
10
12
14
16
Number of handoff call arrivals
Figure 6: Handoff calls accepted (%) Vs. Number of
handoff call arrivals
Figure 7 shows that fuzzy based scheme performs
better in acceptance of new calls for both the cases of
buffer capacity.
% of accepted new calls .Vs. Number of new calls arrivals
% of accepted new calls
0.9
0.8
0.7
0.6
Fuzzy (100 MB)
Non fuzzy (100MB)
0.4
Fuzzy (200 MB)
Non fuzzy (200MB)
0.3
0.2
5
10
15
Number of new calls arrivals
20
[1] Jian Chen, Victor C.M. Leung, ”Applying active
queue management to link layer buffers for real-time
traffic over third generation wireless networks”, Proc.
IEEE Wireless Communications and Networking, New
Orleans, USA, vol. 3, pp. 1657-1662, March 2003.
[2] Yuan Chen, Lemin Li, ”A Random Early Expiration
Detection Based Buffer Management Algorithm for
Real-time Traffic over Wireless Networks”, Proc. 5th
International Conference on Computer and Information Technology (CIT’05), Shangai, China, 2005
[3] Sang-Ho Lee, Kyu-Young Whang, Yang-Sae Moon,
Wook-Shin Han, Il-Yeol Song, ”Dynamic buffer allocation in video-on-demand systems”, IEEE Transactions on Knowledge and Data Engineering, vol. 15,
no. 6, pp. 1535 - 1551, 2003.
1
0.5
such as defense calls, VIP calls, etc., apart from the realtime and non real-time characteristics of an application.
25
Figure 7: New calls accepted (%) Vs. Number of new
call arrivals
4. Conclusions
In this paper, we proposed a scheme for buffer allocation of multimedia applications by using fuzzy logic. The
main objective is to use the base station buffers efficiently
and decrease the call rejections especially for handoff
calls. One important characteristic of our fuzzy based
buffer allocation scheme is that allocation is done looking
at some fuzzy parameters of each application. Parameters
considered are priority, rate of flow, and packet size.
Extensive simulation results reveal that our scheme
features low call rejecting probability, and good buffer
utilization as compared to a traditional buffer allocation
scheme. The scheme can be extended to perform buffer
allocation by considering the criticality of an application
[4] Jingxuan Liu, Ansari, ”Class-based dynamic buffer
allocation for optical burst switching networks”, Proc.
Workshop on High Performance Switching and Routing, Merging Optical and IP Technologies, Torino,
Italy, pp. 295-299, 2002.
[5] Yongho Seok, Jaewoo Park, Yanghee Choi, ”Queue
Management Algorithm for multi-rate wireless local
area networks”, Proc. 14th IEEE Proceedings on Personal, Indoor and Mobile Radio Communications,
Anchorage, Alaska, USA, vol.3, pp. 2003-2008, 2003.
[6] Mats Sagfors, Reiner Ludwig, Michael Meyer, Janne
Peisa, ”Buffer Management for Rate-Varying 3G
Wireless Links Supporting TCP Traffic”, Proc. Vehicular Technology Conference, vol. 1, pp. 675-679,
2003.
[7] Janaki Bandara, ”Resource Allocators for Non Real
Time Traffic in Wireless Sensor Network using Fuzzy
Logic”, International Journal on Wireless Personal
Communications, vol. 21, no. 3, pp. 329-344, June
2002.
[8] T. J. Ross, Fuzzy Logic for Engineering Applications,
1998