Article Preview
TopIntroduction
Nowadays, the IEEE802.11n wireless local-area network (WLAN) has been extensively applied around the world due to its characteristics of the easy installation, the flexible coverage, and the low cost (Crow et al., 1997). WLAN offers the Internet access through the wireless medium. Thus, it consists of several benefits such as mobility, flexibility, and portability. Hence the popularity of WLAN is increasing in offices, schools, hotels, and various public spaces (Dash et al., 2017).
In WLAN, hosts are mostly non-uniformly located (Mittal et al., 2008), and the number of users or the traffics tend to fluctuate unpredictably depending on the time and day of the week (Balazinska & Castro, 2003) (Kotz & Essien, 2005). Furthermore, the conditions of network devices and communication links may be influenced by different factors, such as weather changes, bandwidth controlled by their authorities, power shortages, or device failures (Nadeem et al., 2009).
The authors therefore have studied the elastic WLAN system that dynamically adapts the network configuration depending on traffic demands and network conditions, to reduce the energy consumption and improve the performance (Mamun et al., 2016) (Mamun et al., 2017). In addition, the authors have implemented the testbed using Raspberry Pi as the software access point (AP) using hostapd. Raspberry Pi is a card-size, single-board computer that has the built-in wireless network interface (NIC) supporting IEEE802.11n (Raspberry Pi, 2019).
In WLAN, the fairness of the throughput performance among the hosts is the fundamental request to be satisfied by the network users for the fair network service among them regardless of host locations in the network field (Pilosof et al., 2003). This fair network service is essential for the network service provider. Thus, the fairness issues in WLAN have been extensively explored (Shi et al., 2014)-(Kemerlis et al., 2006). Among them, the TCP throughput fairness is the most important one, because the TCP (Transmission Control Protocol) is the protocol adopted at most of the common network services for users such as Web systems and electronic mails.
However, our preliminary experiments using the elastic WLAN system testbed have revealed that the TCP throughput unfairness will appear among hosts that are concurrently communicating with the same AP, when they are located at different distances from the AP. This unfairness can be caused by the differences in the modulation and coding scheme (MCS) and the TCP congestion window size.
Within our survey, most of existing methods to solve this unfairness problem require the modifications of the standard protocol for data communications in the back-off time algorithm of the CSMA/CA protocol or in the TCP congestion window control. Because they are usually implemented at the operation system (OS) or kernel in a computer, it is necessary to modify the source codes for OS (Pilosof et al., 2003) (Blefari et al., 2007) (Hua et al., 2012) (Freemon, 2014). Then, it causes another serious problem such that every time the OS is updated, this updated OS must be modified to incorporate these methods (Li et al., 2004).