Page 30 - Proceedings of the 2018 ITU Kaleidoscope

2018 ITU Kaleidoscope Academic Conference

Based on the QoE results, the customer satisfaction model 5.2 Unsupervised ML: Wi-Fi patterns and scenarios
(CSAT) described in [14] will estimate the satisfaction of
the user with the service. This constitutes the second “hot For the identification of the different scenarios to be
point” where corrective actions may be necessary based on considered, DBSCAN clustering ML techniques were used.
the detected KRIs that may lead to churn and affect the As a first stage, ML algorithms were applied only to the
business model, as shown in Figure 8. most relevant parameters from the extensive data set
obtained from the probes (bandwidth consumed,
Finally, the KBOs will be updated on a regular basis. As transmission rate, number of data frames, frames in failure,
referred to in [15], the KBOs are derived from the business etc). The results indicated distinct behavior patterns for, at
areas that are determined important for each company and least, three different types of scenarios:
they should be adjusted through the operational efficiency
to increase revenue, reduce cost and improve customer  Commercial & Business scenarios: shopping centers,
experience. This constitutes the last intervention point restaurants, hotels, entertainment venues, etc.
where corrective actions may be required. Billing,  Public scenarios: schools, university campuses,
advertising, fitting the QoS requirements and other cultural venues, museums, etc.
measures should be analyzed to update the SLA, taking into  Residential scenarios: both residential wireless
account the outputs of the user plane and the control planes. access and shared agreement access.

5. CASE STUDY: IEEE 802.11 SCENARIOS Therefore, it was decided to carry out the study in three
such real Wi-Fi scenarios:
In previous sections we have introduced the principles of  The CHQ building (https://chq.ie/): to cover both
the QoE enhancement methodology approach. In this business and commercial scenarios. This building,
section, we present the implementation of the proposal in a situated within the heart of Dublin city, holds
real next-generation wireless scenario: the Wi-Fi networks. different food and shopping spaces.
 Dublin Institute of Technology (http://www.dit.ie/):
5.1 Wi-Fi technology to cover public (e.g. educational campus) scenarios.
 The Gasworks Area: to cover residential scenarios.
Wi-Fi technology, also known as IEEE 802.11, was
originally designed to be a wireless local area network Subjective and objective measurements were required to
access technology which is to say intended only for small validate the methodology. Objective Wi-Fi information was
coverage areas. However, Wi-Fi has become a “ubiquitous collected through OptiWi-fi network probes described in
access technology for mobile users” [1]. Consequently, it is [16]: NP data, location data, mobility patterns, number of
now being considered as one of the key access technologies access points (AP), number of clients, connection time, type
in the 5G HetNet ecosystem. Nevertheless, QoS of devices and other information, according to [17], was
management in this type of “unlicensed radio spectrum” captured by the probes. Subjective information was
wireless network may become very complex, especially gathered through user surveys: personal information
when users' demands in terms of QoE are increasing. (gender, age, occupancy, Internet expertise) together with
the data related to the “Wi-Fi experience”, both about the
IEEE 802.11e was developed to offer new QoS capabilities user’s requirements/expectations and QoE/satisfaction with
to the IEEE 802.11 WLAN networks. Under this the service, were collected in the same survey. The dates of
amendment, new QoS mechanisms, such as the enhanced the test field in each of the scenarios are presented in Figure
distributed channel access (EDCA) mechanism were 9 where the surveying process was carried out in a one-
introduced to Wi-Fi technology at the media access control week period for each venue due to the complexity of the
(MAC) layer, enabling different classes of access categories procedure and at least 50 participants took the survey in
(AC) in order to support the prioritization of distinct classes each of the scenarios. The network probes were deployed
of services. Even though numerous research studies can be for a month in each of the venues to capture the required
found for the implementation of this protocol in scientific information about context and system influence factors.
literature, it is rarely used in real deployments. The reason
is quite obvious: Wi-Fi QoS management is completely
dependent on the behavior of users, the coexistence of
networks and many other aspects that the provider cannot
control. That is the reason why most Wi-Fi service
providers currently resort to an "over-dimensioning" in the
deployment of their access points (AP).
For the above, the Wi-Fi RAN has been selected to validate
the proposed methodology to enhance QoE.

Figure 9. Test plan

– 14 –