Page 242 - Big data - Concept and application for telecommunications
P. 242

5                                Big data - Concept and application for telecommunications



            6.8     Big-data-driven resource management

            Network operators should be aware of their long-term deployment objectives in terms of network capacity,
            coverage, and the number and locations of base stations. They also need new resource allocation strategies
            in order to fulfil different traffic demands or requirements across the entire coverage area. To achieve these
            goals, operators have been monitoring network QoS through driving tests with smartphones. Measurement
            results are gathered from selected smartphones or specific driving test phones in their networks, which are
            analysed by specialized software. However, this is not cost-effective due to excessive consumption of time
            and human resources, and is also inaccurate due to the limited number of test samples.
            Thus, the use of big-data analytics can provide a new way to tackle these problems. Network analytics involve
            monitoring and analysing real-time and history data across users, mobile networks and service providers.
            Network resources can be dynamically allocated in real time by analysing network resource data, traffic data
            and user information.

            By utilizing data analytics, changing resource requirements from one location to another in a specific period
            become predictable. In addition to network data, behavioural and sentiment analyses from social networks
            and other sources require consideration to predict where and how users may use the mobile network. For
            example, when a social event such as a marathon takes place in a city, some places like the streets on the
            race route may attract large crowds of people, resulting in potential congested traffic in these locations
            during the event. Hence, with this predicted information from data analytics, operators can allocate more
            radio  resources  to  the  hotspot  in  such  a  way  that  the  peak  traffic  can  be  absorbed  smoothly  without
            sacrificing user QoE.

            Users often travel from one place to another around the city (e.g., work in the central business district during
            the day and live in a suburb at night). This causes the traffic of each cell to fluctuate significantly at different
            times  of  day,  which  is  dubbed  the  ''tide  effect.''  If  resources  are  allocated  to  each  cell  with  a  fixed
            configuration, resource utilization must be underestimated, and it is difficult for users in the hotspot to obtain
            good QoE during peak hours. On the other hand, a great deal of resources may be wasted in idle times at low
            traffic locations. Current and historical data can be utilized by data analytics to predict traffic for high-density
            areas in the networks. Then, with the radio access network (RAN) architecture, predictive resource allocation
            in centralized baseband units may help to accurately serve the right place at the right time (i.e., knowing
            when and where peak traffic arises), causing minimum disruptions to services.

            6.8.1   Big-data-driven resource allocation based on user interest
            By evaluating closeness (both geographic and social) among users in the same base station, users with high
            closeness are categorized into a cluster. Users in the same cluster can share a wireless channel and receive
            the  same content, which not only  improves  the  data  rate  for  cluster  users,  but  also saves  base  station
            resources.

            Figure 6-7 depicts the use case of big-data-driven resource allocation based on user interest.






















                            Figure 6-7 – Big-data-driven resource allocation based on user interest


            234      Network and infrastructure
   237   238   239   240   241   242   243   244   245   246   247