ATM Switch: Impact Of Heuristic Approach To Static Bandwidth Allocation

Wee Kuokkwee

, M. Othman, S. Shamala, and A.A Ariffin

ATM Switch: Impact Of Heuristic Approach To Static
Bandwidth Allocation


1Wee Kuokkwee, 2M. Othman, S. Shamala, and A.A Ariffin

Department of Communication Technology & Network
Faculty of Computer Science & Information Technology
43400, University Putra Malaysia, Serdang,
Selangor Darul Ehsan, Malaysia.
Tel: 603-8946 6565
Faks: 603-8948 2102

1kuokkwee@yahoo.com
2mothman@fsktm.upm.edu.my


Abstract
average delay are the evaluation parameters

in ATM QoS requirement that assess a
In this paper, the performance bandwidth strategy’s performance. Although
evaluation of two Asynchorous Transfer network performances in terms of cell loss
Mode (ATM) scheduler strategies, static ratio and average cell delay from [3] are
bandwidth allocation with and without better for bandwidth allocation strategy
heuristic approach are presented. Several called Bandwidth Allocated Proportional to
experiments were carried out by using the Expected Queue Length with Threshold
ON-OFF distribution source model based on Value, this is true only when the buffer space
the ATM Various Bit Rate (VBR) service exceed 500 cells. For real-time applications,
category. The strategy with the heuristic generally these types of traffic cells are not
approach improves the buffer cell loss ratio queued in the buffer, as this would introduce
(CLR) performance significantly. Several cell delay. However, if traffic shaping is
numerical results are presented to show the required, cells may need to be buffered and
effects of the heuristic approach to large buffer space is better to avoid due to
bandwidth allocation.
their stringent delay requirement [4].

Bandwidth allocation is defined as
Keywords: ATM, Bandwidth Allocation,
allocating bandwidth to all the connections at
VBR, Heuristic Approach.
contention point, which usually is called a

switching unit. In order to increase network
1. Introduction
throughput, bandwidth needs to be managed

and allocated fairly and this is the
Real time video, such as video responsibility of the ATM switch. Two
conferencing and video-on-demand, is bandwidth allocation strategies have been
expected to be one of the major services developed in this research. They are static
supported by the ATM-based ISDN [1]. The bandwidth allocation from [2] and an
main service used for transmitting video enhancement strategy with heuristic
streams is variable bit rate (VBR) approach.
transmission. Cell loss ratio (CLR) and cell

32

---

ATM Switch: Impact Of Heuristic Approach To Static Bandwidth Allocation

Only the results for VBR Video/Data ATM networks and so takes care of cell
buffer and VBR Video traffic are given in scale congestion. This buffer is the limited
the discussions. The comparison is done to resource, which is critical to the
evaluate the performances of static performances of the simulation model. DTS
bandwidth allocation, which adopts the model is where several bandwidth allocation
heuristic approach and the one without strategies operate and where bandwidth is
heuristic approach.
calculated and allocated to the buffers. A

total number of two million time slots are
2.
System Description and Model executed for the entire duration of the
Development
simulation.

Static Bandwidth Allocation is a
The simulation model utilized from this strategy that guarantees bandwidth to high
research is shown in Fig 1. Two bandwidth priority traffic classes. This allocation does
allocation strategies have been developed; not change with time. Thus, it provides no
they are Static Bandwidth Allocation and congestion control. In this study, VBR
Static Bandwidth Allocation with heuristic Video/Data buffer is allocated 300 Mbps,
approach. The first strategy is taken from [2],
Connectionless Data buffer is allocated 20
while the second strategy is the modification Mbps and VBR Video buffer is allocated 300
of previous strategies implemented in [2] Mbps. This pattern of bandwidth allocation
with the purpose to improve the network is fixed for every cycle length. Thus, this
performances.
strategy does not have bandwidth allocation

recalculation mechanism. The amount of
Time
bandwidth allocated to buffer is calculated as
Slot
Buffer
below:

622
VBR Video/Data Bandwidth Allocated
Dynamic
Mbps
Time
= 300Mbps * 200 Slots.
Slice
Server

622Mbps

Connectionless Data Bandwidth Allocated


= 20Mbps * 200 Slots.
Fig 1: Simulation Model
622Mbps


To investigate the proposed algorithm VBR Video Bandwidth Allocated
and its efficiency, a single virtual path with

= 300Mbps * 200 Slots.
622 Mbps is investigated. Typically there 622Mbps
could be diverse services using this path and

so a number of source models are needed.
The Static Bandwidth Allocation with
The ON-OFF source model is used to Heuristic approach is an enhancement of the
represent some of the VBR services in the previous strategy [3] and only implemented
ATM networks, which are VBR Video, for the highest priority traffic type. A new
Connection Oriented Data, VBR Video/Data buffer called Dual Leak Bucket (DLB) is
and Connectionless Data. Source model is introduced. The DLB consists of two buffers,
where the ATM cells are generated. The cell primary and secondary. When the primary
stream from a number of sources is then queue length exceeds the capacity of the
input to the ATM switch buffers model. The buffer, the following cells will be inserted
buffer smoothes the arrival of cells to the into the secondary buffer based on FIFO
International Journal of The Computer, the Internet and Management Vol. 16. No.1 (January-April, 2008) pp 32-37


33

---

Wee Kuokkwee, M. Othman, S. Shamala, and A.A Ariffin

discipline. The cells in secondary buffer will 3. Results and Discussion
be transferred to the primary queue once

there is an empty slot. To minimize the
In this experiment, the buffer sizes for
system overhead, cells in secondary buffer VBR Video/Data and VBR Video are varied
will be copied into primary buffer based on from 200 cells to 1200 cells, whereas
three criteria. The first criterion is at the inter
Connection Oriented Data infinite buffer size
cycle period. In other words, it is defined as and Connectionless Data remains 100 cells.
the time when a cycle has finished and prior The output parameters are cell loss ratio and
to entering into the next cycle. Second, when
average cell delay. The network performance
the secondary buffer is full. Third, there is no
has been plotted out illustruting the output
cell to be transmitted in the primary buffer.
parameters versus buffer size. The
The heuristic approach algorithm is simulation was carried out on a Pentium III
elaborated in Fig 2:
1000 MHz system using Visual C++ 6.0.

The cell loss ratio performances in VBR
/*decide which buffer to insert the cell*/
Video/Data buffer for these two types of
Choose_Buffer(){
static bandwidth allocation strategies are
//check buffer size
depicted in Fig 3. It is obvious that the static
If (queue[index] < buffer_size[index]){
bandwidth allocation strategy with heuristic
Copy();
// copy cells to primary
approach has a better cell loss ratio
//switch back to primary buffer if no cell in
performance than the traditional static
//secondary buffer
bandwidth allocation strategy overall. There
If(queue[0] < buffer_size[0] & queue[4] = = 0)
are differences of cell loss ratio between the
// 0 denotes primary buffer
static bandwidth allocation with heuristic and
Index = 0 // 4 denotes secondary buffer
without heuristic approach when the buffer’s
else


size is small, 0.0120753 and 0.0051269 for
Index = 4
the VBR Video/Data traffic when the buffer
}
is set to 200 and 400 respectively. The
}
traditional static bandwidth allocators

produce higher cell loss ratio, which is equal
/*copy cells to the primary buffer */
to 0.0120753 and almost double of the cell
Copy(){
loss ratio as compared to the bandwidth
If( buffer size > queue length){
allocator with heuristic approach when the
If((buffer size - queue length) > secondary queue
buffer size is 200 which is shown in Fig 3.
length)
However, the differences between these two
Copy all cells from secondary queue
strategies become smaller when the buffer
Else
size is increased.
Copy cells based on the empty space in
Although the cell loss ratio for VBR
primary buffer
Video Buffer is always within the range of
}
0.018 till 0.025 for both strategies, the
else /* do nothing */
strategy with heuristic approach performs
}
better than the traditional strategy. From the

Fig 4, it can be concluded that the static
Fig 2: Heuristic Approach In Bandwidth bandwidth allocation produces the more
Allocation
consistent cell loss ratio.

The comparison of the average cell
delay performance in VBR Video/Data
buffer for both strategies is illustrated in Fig

34

---

ATM Switch: Impact Of Heuristic Approach To Static Bandwidth Allocation

5. It is observed that the average cell delay both strategies becomes smaller as the buffer
for strategy with heuristic approach is always
size increases.
higher. This is caused by the cells of the
Average cell delay for the VBR Video
strategy with heuristic approach that require Buffer performances in the static bandwidth
to queue and wait before they can be allocation with heuristic approach and
transmitted when the scheduler is congested. traditional static bandwidth strategy which
As shown in Fig 5, when the buffer size is are shown in Fig 6 are similar to the average
200, the cell delay of the strategy with cell delay for VBR Video/Data buffer. As
heuristic approach is 0.000198, which is mentioned earlier, this is due to the large
0.000085 seconds longer than the strategy number of VBR sources in the network.
without heuristic. However, the delay for

VBR Video/Data Buffer's Cell Loss Ratio
0.025
0.02
i
o
0.015
0.01
l
Loss Rat
0.005
Cel
0
200
400
600
800
1000
1200
1400
Heuristic
0.0078609
0.001165535
4.77E-04
0.00018746
0
0
0
Without Heuristic
0.0199362
0.00629244
0.00257885
0.000837966
5.96E-05
7.04E-05
0

Fig 3: VBR Video/Data Buffer’s Cell Loss Ratio
VBR Video Buffer's Cell Loss Ratio
0.026
tio
0.024
a
R
0.022
s
s
o
0.02
ll L
e
0.018
C
0.016
200
400
600
800
1000
1200
Heuristic
0.0201492 0.0196404 0.0187658 0.0188531 0.0195143 0.0185326
Without Heuristic 0.0245634 0.0196663 0.0214342 0.0202302 0.0207348 0.018815

Fig 4: VBR Video Buffer’s Cell Loss Ratio
International Journal of The Computer, the Internet and Management Vol. 16. No.1 (January-April, 2008) pp 32-37


35

---

Wee Kuokkwee, M. Othman, S. Shamala, and A.A Ariffin




VBR Video/Data Buffer's Average Cell Delay
0.0004
0.00035
ay
el
0.0003
l
D
el
onds)
0.00025
ec
0.0002
age C
(
S
ver
0.00015
A
0.0001
200
400
600
800
1000
1200
1400
Heuristic
0.000198 0.000285
0.0003
0.000302 0.000304 0.000322 0.000365
Without Heuristic 0.000113 0.000195 0.000244 0.000255 0.000258 0.000301 0.000343


Fig 5: VBR Video/Data Buffer’s Average Cell Delay

VBR Video Buffer's Average Cell Delay
0.00175
l
ay
0.0015
e
0.00125
s)
l
l
D
e
0.001
0.00075
e C
econd
ag
0.0005
(
S
0.00025
ver
A
0
200
400
600
800
1000
1200
Heuristic
0.0001995 0.0004731 0.0007219 0.0009882 0.0013888 0.0016351
Without Heuristic 0.0001985 0.0004527 0.0006999 0.0009513 0.0012016 0.0014286


Fig 6: VBR Video Buffer’s Average Cell Delay


36

---

ATM Switch: Impact Of Heuristic Approach To Static Bandwidth Allocation

4. Summary
Performance Modeling with Dynamic

Bandwidth Allocation”. Malaysian
The analysis of all graphs proves that a
Journal of Computer Science, Vol. 2,
total amount of 400 buffer cells and above
pp. 79-86.
are suitable for cell loss ratio, whereas 800
buffer cells and above will produce a more [4] M. Othman, C.H. Loh, A.K.Ramani,
stable average cell delay. Judging from the
D.Shyamala and L.N.Abdullah (2001).
performances evaluation results, it can be
“Dynamic Bandwidth Allocation with
concluded that the static bandwidth
Low Buffer Storage in ATM Switch”.
allocation with heuristic approach is more
Elektrika 4, Vol. 2, pp. 62-67.
efficient than the traditional static bandwidth
allocation strategy. Finally, it can be [5] Ossama M. Younis, Wafaa A. El-
observed that network performances are
Haweet, M. Nazih El-Derini (1999).
highly dependent on the bandwidth
“Dynamic Bandwidth Allocation for
allocation strategies.
VBR Video Sources with Similar or

Different QoS Requirements in ATM
References
Networks”
in
IEEE MiddleEast

Workshop on Networking, Lebanon.
[1] H.J. Stuttgen.(1995). “Network
Evolution and Multimedia
[6] Piya Tanthawichian, Akihiro Fujii,
Communication”. IEEE Multimedia,
Yoshiaki Nemoto (1998). “Bandwidth
Vol. 2, pp. 42-59.
Allocation in ATM Networks:

Heuristic Approach. Compute
[2] Lu Wei, Peter Hansson. (1996).
Communications and Network” in
“Further Research On Dynamic Time
Proceedings 7th International
Slice in ATM Switch”, Computer
Conference, pp. 20-25.
Networks and ISDN Systems, Vol. 28,
pp. 789-798.
[7] S. Dixit and P. Skelly (1995). “MPEG-

2 over ATM for Video Dial Tone
[3] M. Othman, Loh C. H., Shyamala D.,
Network: Issue And Strategies”. IEEE
L.N. Abdullah. (1999). “ATM Switch
Network, Vol. 9, pp. 30-40.


______




International Journal of The Computer, the Internet and Management Vol. 16. No.1 (January-April, 2008) pp 32-37


37

---

Document Outline

• ??
• ??
• ??
  • ??

---