Achieving Load Balancing in SDN Based High Density WiFi Network

Due to simple technical implementation and low costs of WiFi networks the wireless devices and applications have increased resulting in a significant increase in internet traffic. The increased traffic load overburdens the network resources resulting in uneven load balancing . In order to improve th

Project Title

Achieving Load Balancing in SDN Based High Density WiFi Network

Project Area of Specialization

Information & Communication Technology

Project Summary

Due to simple technical implementation and low costs of WiFi networks the wireless devices and applications have increased resulting in a significant increase in internet traffic. The increased traffic load overburdens the network resources resulting in uneven load balancing . In order to improve the network utilization and to maintain the quality of service (QoS), load balancing is important in high-density WiFi networks which is a difficult task in traditional WiFi networks where association and de-association decisions are made by the wireless devices .

Recently a new paradigm software defined network (SDN) is introduced that has the functionality to manage, control and measure high density Wi-Fi networks . The control plane allows the administrator to manage the network by having an advantage of the overall view of the network. SDN provides the flexibility to alter the traditional WiFi networks through programmability and without changing the hardware hence supporting the load balancing algorithms. The control plane communicates to the data plane or forwarding devices through OpenFlow protocol.

To solve the problem of WiFi obstruction among the OpenFlow enabled APs (OAPs), we recommend a QoS-aware load balancing strategy (QALB) for software defined Wi-Fi networks (SD-Wi-Fi). The SDN controller selects a load level up to which OAPs make the association and de-association decisions themselves. The wireless devices from an overloaded OAP are shifted to the destination OAP by considering multi-metrics such as packet loss rate, throughput and received signal strength indicator (RSSI). We conduct our comparative experiments using emulation tools (Mininet-NS3-WiFi) and a Linux based test bed with same configurations to measure the performance evaluation of the high density software defined WiFi networks (SD-WiFi).

Project Objectives

In this research we propose a QoS-aware load balancing strategy for high density Wi-Fi networks by taking advantages of the SDN to improve the overall network performance. The SDN controller harvesting an overall view of the network helps in collecting information from the OAPs and decide up to which load level the OAPs can carry out the association decisions by themselves, without consulting the centralized controller. There exists some literature that make use of the dedicated centralized controller to carry out the load balancing functions in the Wi-Fi networks. Our solution brings novelty in three ways. The existing load balancing methods relying on a centralized approach, renounce all the AP, RSSI based association and de-association decisions to the centralized controller. Our emulation and testbed results show that 100 percent dependency on the centralized controller increases the turnaround time for AP association and de-association process and further overloads the controller. The proposed strategy allows the OAPs to make the association decisions themselves by dynamically adjusting the load level according to the network conditions. The second aspect of the proposed study is that, the destination OAP is selected through a multi-metric criterion (packet loss rate, RSSI and throughput), satisfying the least loaded conditions instead of just relying on one metric i.e., RSSI. Finally, our solution does not require any hardware changes and is applicable to any wireless device which supports the OpenFlow standards.

Project Implementation Method

The OAPs boot or reboot subject to the condition when the OAPs are operational. The OAPs communicate their load and capacity information to the SDN controller via OpenFlow session, whenever a status change occurs at the OAPs. If the status of the OAP is not changed, the status is rechecked after 10 seconds. If the status of the OAP changes means the OAP receives a new association request or the load of the OAP varies to the previous calculated load, the OAP reports the load and capacity information to the SDN controller. The controller on receiving the load and capacity information calculate the load levels. If current wireless stations (cw) associated to a certain OAP exceeds its suggested number of wireless stations (sw), the controller issues a de-association list of wireless stations to the overloaded OAP. The OAP de-associates the wireless stations with the weakest RSSI connections. The underloaded OAPs with the best communication quality are issued a re-association list by the controller to re-associate the wireless stations which were de-associated earlier on. In this way the load is optimized among the OAPs and wireless stations connect to only those OAPs having the best, maintaining the QoS at the users end. The process is repeated periodically after 10 seconds to ensure the fairness of load among the OAPs. 10 seconds provide enough time for the mobile wireless stations to associate, de-associate or re-associate to the neighborhood OAPs in the emulation environment.

The SDN controller collects the load and capacity information from the OAPs. On receiving the information, the SDN controller computes the number of BSSs in the ESS (nb). If nb = 1 then the load balancing condition is met. If there are more than one BSSs in the ESS the controller computes the load levels. If J is computed to be 1, this means the load among the OAPs is balanced. On the other hand, if J is not equal to 1 then the SDN controller prepares a de-association eassociation list of the wireless stations for the OAPs. The OAPs load is computed by using the packet loss rate, RSSI and the normalized throughput values. The OAPs having load levels greater than the Lavg receive the de-association list of wireless stations having the weakest RSSI connections and the OAPs having load less than the Lavg receive the re-association list to re-associate the de-associated wireless stations.

A number of evaluation platforms exist to perform experimentation on wired networks instead of the wireless networks. High density SD-Wi-Fi require an urgent need for such platforms, that can support the performance evaluation. We first emulate and then design a prototype testbed to evaluate the performance of the proposed QoS-aware load balancing strategy in high density SD-Wi-Fi.

Benefits of the Project

The past centralized WiFi load balancing methods push all the AP association/re-association decisions to the centralized controller. Our emulation and testbed results show that by pushing all the association/re-association decisions to the centralized controller increases the turn around time for the AP association and de-association process and further overloads the controller. Our proposed methods setups a load level dynamically according to the network condition, up to which all the OAPs association and re-association decisions are carried by the OAPs themselves.

In the WiFi load balancing methods, the wireless stations associate to the APs with the best RSSI strength. Associations made on only RSSI parameter create crowded hotspots and degrade the throughput performance. In the proposed strategy, the destination OAP is selected based on multi-metric criterion and not only on the RSSI strength. Packet loss rate, RSSI and throughput are considered as the main factors to decide whether an OAP is the best candidate to re-associate to. The multi metric access selection allows improved network performance in terms of throughput, average frame delays and average number of retransmissions.

The WiFi load balancing methods either tame client side or AP side to achieve fairness among the network resources. Our proposed strategy takes into account both the client side and AP side information in order to achieve load balancing without changing the hardware. We have modified the standard OpenFlow messages to incorporate necessary fields which support the QoS-aware load balancing. The proposed QALB is applicable to all the wireless devices that support the OpenFlow standards.

Technical Details of Final Deliverable

QALB distinguishes itself from the existing methods of load balancing that used a centralized approach, in three aspects. The overloading of the centralized controller is alleviated in QALB, by a load level, that is used to offload the OAPs associations and de-associations from the controller to the OAPs. Secondly the minimum value of the load level is calculated automatically, which helps in maintaining the load balancing among the OAPs. The third aspect involves the OAP selection, based not only on RSSI but other important QoS factors such as packet loss rate and throughput. The OpenFlow allows the OAPs to report the capacity and load information to the SDN controller, whenever they boot e-boot or their status change. The status change reflects the occurrence of associationde-association or the load variation of an OAP. After receiving the OAPs information, the SDN controller computes the fairness index. If the fairness index is not close to 1, the SDN controller issues the load balancing instructions to the OAPs. The overloaded OAPs are instructed to de-associate the wireless stations with the weakest RSSI connection. The wireless stations that get de-associated will try to associate with another OAPs. Only the specific underloaded OAPs chosen by the SDN controller will respond to the association requests.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Telecommunication

Other Industries

Core Technology

Others

Other Technologies

Sustainable Development Goals

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com