Analysis on the effect of first-in-first-out queue length ratio to the multi-hop wireless network performance
A multi-hop wireless network is created by connecting multiple wireless access points as the backhaul of the network to increase the network coverage. The issue of spatial bias, unbalanced network performance of end-to-end throughput and delay occurs when the total offered load of the associated sta...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/14752/1/Analysis%20On%20The%20Effect%20Of%20First-in%20First-out%20Queue%20Length%20Ratio%20To%20The%20Multi-hop%20Wireless%20Network%20Performance%2024%20pages.pdf http://eprints.utem.edu.my/id/eprint/14752/2/Analysis%20on%20the%20effect%20of%20first-in-first-out%20queue%20length%20ratio%20to%20the%20multi-hop%20wireless%20network%20performance.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A multi-hop wireless network is created by connecting multiple wireless access points as the backhaul of the network to increase the network coverage. The issue of spatial bias, unbalanced network performance of end-to-end throughput and delay occurs when the total offered load of the associated stations exceeds the wireless link capacity. Station associated
to the access point more hops away from the gateway will experiences significant amount of delay and lower end-to-end throughput compared to the station fewer hops to the gateway. To demonstrate the issue of spatial bias, a Linux based multi-hop wireless network testbed was constructed with six mesh access points (MAP) and a mesh portal. The MAP consists of two ingress interfaces (one to allow the association of station (local
ingress interface) and one to allow other MAP to associate to it (mesh ingress interface)) and one egress interface to associate to another MAP. The wireless link capacity of the constructed testbed is determined by the amount of offered load that is about to congest the network. A non-congested access point has the sum of the arrival rate of both the mesh and
local ingress interface not larger than the wireless link capacity. Every packet received by both the ingress interfaces of a non-congested access point will be almost immediately forwarded (packets will stay in the transmit queue awhile due to the processing delay) to the destination. However, packet received by a congested access point will be competing not to be dropped and subsequently enqueued into the transmit queue successfully. A transmit buffer (queue of waiting packets) is commonly allocated to the egress interface to fully utilize the wireless link capacity. The process of enqueueing packets into the transmit buffer is handled by a queueing manager (First-In First-Out is the queueing discipline used by the Linux queueing manager). The equality of local successful transmit probability (an) and mesh successful transmit probability (b,,) in congested MAPs, which is the main root cause of the spatial bias problem, is modelled and validated. The proposed solution for the
spatial bias problem is to allocate individual transmit buffer with different successful transmit probability for the two ingress interfaces. The hypothesis, "the ratio between the length of local and mesh ingress interface queue can affect the successful transmit probability of the respective interface" is validated by three queueing configurations, namely LIOO_M500, LIO_M50 and LlO_M40 that have queues with different length ratios
in congested MAPs. If packet arrival ratio of local over mesh ingress interface is larger than the respective queue length ratio, the mesh ingress interface successful transmit probability will be higher than the local ingress interface successful ransmit probability. On the other hand, if packet arrival ratio of local over mesh ingress interface is smaller than (or equal to) the respective queue length ratio, the mesh ingress interface successful transmit probability will be lower than (or equal to) the local ingress interface successful transmit probability. The effect to the end-to-end throughput and delay introduced by the proposed solution is analysed. By controlling the ratio of queue lengths, the spatial bias problem in multi-hop wireless network can be alleviated. |
|---|