It is studied that network dynamics such as link hysteresis and node mobility in wireless ad-hoc environments lead to frequent communication failures. Due to that, routing overhead is a major hindrance to configuring real-time application, such as VoIP, in ad-hoc environment, because control packets may impose severe quality of service (QoS) degradation. Therefore, our objective is to understand the impact of varied packet sizes and hop-counts in a real-time ad-hoc network by setting-up an ad-hoc testbed. Specifically, our objective is to study how routing overhead, packet size and hop-count interact with each other, and to what extent they affect end-to-end delay and throughput. Therefore, we aim to obtain real-time measurements in different scenarios, and by using those results we aim to develop a statistical-analytic approach to studying the effects of routing protocols on delay and throughput.

The real-time testbed consists of various mobile devices, such as iPAQ, Sharp Zaurus, laptop and desktop machines each equipped with wireless cards, consisting different hardware and software platforms. The different types of hardware devices helps us obtaining real-time measurements which will be applicable to heterogeneous environment as well. All wireless nodes communicate in ad-hoc mode, and data streams with varied packet sizes and the testbed topologies consisting of varied number of hops are used to generate several load conditions in the experimental setup. Although, the testbed can be used to evaluate any ad-hoc routing protocol, as a case study, we study optimized link state routing (OLSR) protocol and ad-hoc on demand vector routing protocol (AODV).

In ad-hoc networks, link failures are generally due to node mobility and link hysteresis. To emulate link failure behavior in the indoor environment, intermediate nodes along the paths are switched on and off once every 10 seconds during data transmission among nodes. In our experiments, every scenario consists of an alternate path to continue data transmission in case of link failure events. The IPtables module is configured to achieve physical separation among nodes that are more than one-hop away by using their medium access control (MAC) addresses. In addition, the transmission rate for each wireless card has been set to 11 Mbps. Access point ID on all nodes is set to a unique value to enable them communicate in ad-hoc mode. The wirelesschannel shared by all nodes are set to 10, whereas the campus network uses wireless channel 1, to avoid interference from the campus wireless network.

Hardware

• Desktop: Dell PC, Pentium IV 2.6 GHZ (RHL 9, kernel version 2.4.20).
• iPAQ: Intel StrongARM 206 MHx (Familiar Linux 0.6, kernel version 2.4.18-rmk3).
• Sharp Zaurus: Intel XScale 400 MHz (Linux Embedix kernel version 2.4.18-rmk7-pxa3-embedix).
• Dell Laptop MN: Celeron Processor 2.4GHZ (RHL 9, kernel version 2.4.20).
• Wireless Cards: Netgear MA 311 802.11b, Lucent Orinoco Gold 802.11b. 

Software

• Kernel AODV version 2.1 from NIST.
• OLSR version 0.4.5.
• IPtables for filtering packet at MAC level.
• Ethereal network packet analyzer.
• Netperf, ttcp, Rude data transmission utilities.

• OLSR
• AODV
• Familiar Linux
• IPtables
• Netperf
• Ethereal
• hping (Another ping utility)
• iPerf (TCP traffic generator)
• Rude (UDP traffic generator)
• Orinoco device drivers
• iPAQ device drivers
• Sharp zaurus device drivers

1. Avesh Agarwal and Wenye Wang, "Statistical Analysis of the Impact of Routing in MANET Based on Real-Time Measurements," in Proc. of the 14th International Conference on Computer Communication and Networks (ICCCN'05), San Diego , California , USA , October, 2005. (Nominated for Best Paper Award)