icc2012 (slides pptx)

Report
Presented at ICC 2012 – Wireless Network Symposium – June 14th 2012
Motivation
 Background
 Experiment Setup & Methodology
 Performance Evaluation
 Discussion
 Conclusions & Future Work

ICC 2012- Wireless Network Symposium - June 14th, 2012
2

Modern devices have
many options for
connectivity (radio
access technologies) or
RATs
ICC 2012- Wireless Network Symposium - June 14th, 2012
3

How can we seamlessly
switch between RATs or
enable multiple RATs
simultaneously?
ICC 2012- Wireless Network Symposium - June 14th, 2012
4

Interaction between
RATs is not well studied

Simulation tools often
ignore interference
between Wi-Fi and
Bluetooth for example
ICC 2012- Wireless Network Symposium - June 14th, 2012
5

Interaction between
RATs is not well studied

Certain algorithms
assume homogeneous
link capacities
ICC 2012- Wireless Network Symposium - June 14th, 2012
6
Source: Wikipedia
Source: Cisco
ICC 2012- Wireless Network Symposium - June 14th, 2012
7

Many reasons why the ability to seamless
switch RATs is important:
› More robust and reliable communications
 Failover to other networks using different RATs
› Potentially increased battery life
 Selecting the most energy efficient RAT when
possible
› Potentially improved performance
 Enabling multiple RATs at once
ICC 2012- Wireless Network Symposium - June 14th, 2012
8

Previous studies show that partial
overlapping of Wi-Fi channels does not
significantly affect performance
[Mishal et al, 2006]

What about overlap between RATs(ex:
Wi-Fi, Bluetooth, Zigbee)?
ICC 2012- Wireless Network Symposium - June 14th, 2012
9

Will performance degrade with a small
channel interfering with a larger
channel? (Bluetooth interfering with Wi-Fi)
› If not, is there a threshold number of
Bluetooth interferers that causes
performance degradation?
ICC 2012- Wireless Network Symposium - June 14th, 2012
10

Will performance degrade with a larger
channel interfering with a smaller
channel? (Wi-Fi to Bluetooth interference)
ICC 2012- Wireless Network Symposium - June 14th, 2012
11

Shuaib et al [2005]
› Bluetooth  Wi-Fi  Internet
 Wi-Fi performance somewhat dependent on
Bluetooth performance
 Also only looks at Wi-Fi performance
 We aim to send Bluetooth and Wi-Fi traffic
independent of each other simultaneously
 We aim to investigate both Wi-Fi and Bluetooth
performance
ICC 2012- Wireless Network Symposium - June 14th, 2012
12

Carfang et al [2008]
› More focused on how mobility affects
communication in the face of interference

Guo et al [2010]
› Focused on error rates and signal strengths
while we are concerned with throughput
ICC 2012- Wireless Network Symposium - June 14th, 2012
13
ICC 2012- Wireless Network Symposium - June 14th, 2012
14

In Bluetooth 1.2: Adaptive Frequency
Hopping
› Tries to sense what channel Wi-Fi is using and
avoid those frequency ranges
› What happens when Wi-Fi saturates the area as
is common now?

Other techniques where Wi-Fi and
Bluetooth chips communicate directly to
avoid
› What happens when using external Bluetooth
adaptors, chips that do not support etc?
ICC 2012- Wireless Network Symposium - June 14th, 2012
15

An equal number of
Bluetooth and Wi-Fi
devices

Located within a lab
at the University of
Guelph
ICC 2012- Wireless Network Symposium - June 14th, 2012
16
ICC 2012- Wireless Network Symposium - June 14th, 2012
17
FTP file transfers between nodes
 Input from /dev/random on Ubuntu 11.10
 Wi-Fi file size was 10 megabytes
 Bluetooth file size was 1 megabyte


Difference is because of the order of
magnitude difference in link capacity so
that a single transfer would finish roughly
at the same time
ICC 2012- Wireless Network Symposium - June 14th, 2012
18

Up to 6 laptops using each technology (6
Bluetooth and 6 Wi-Fi laptops)

Experiment performed late in the evening
when few students use the campus Wi-Fi to
decrease external interference
30 Repetitions for each data point
 Throughput is averaged across all nodes
using particular technology

ICC 2012- Wireless Network Symposium - June 14th, 2012
19
ICC 2012- Wireless Network Symposium - June 14th, 2012
20
ICC 2012- Wireless Network Symposium - June 14th, 2012
21

Bluetooth performance particularly is
affected by Wi-Fi nodes in the region
› Fig 2: 2, 3, 4 devices all show decreased
performance due to interference

In some cases Wi-Fi nodes are also
affected by Bluetooth interference
› Fig 3: 3 devices
ICC 2012- Wireless Network Symposium - June 14th, 2012
22

Decreased performance due to
increased contention within the
technology more to blame than
interference between technologies, but
still some effect
 In future experiments, it may be best to vary
the number of one type of node at a time to
isolate the cause of decreased performance
ICC 2012- Wireless Network Symposium - June 14th, 2012
23

Results show that Bluetooth is significantly
affected by Wi-Fi interference within the
same channel range

Result also showed limited cases of Wi-Fi
being affected by Bluetooth interference

Results make the case that simulation tools
should begin to support interference
modelling between technologies
ICC 2012- Wireless Network Symposium - June 14th, 2012
24

In future experiments, vary number of nodes
with a particular RAT while leaving other
constant
Expand to include other RATs, ex Zigbee
 Vary the distance, introduce mobility


Develop heterogeneous access schemes
that avoid interference between RATs
ICC 2012- Wireless Network Symposium - June 14th, 2012
25
Questions?
Jason Ernst, University of Guelph
[email protected]
ICC 2012- Wireless Network Symposium - June 14th, 2012
26

similar documents