Page 247 - Big data - Concept and application for telecommunications
P. 247
Big data - Concept and application for telecommunications 5
6.12 Big-data-driven content delivery network deployment
The content delivery network (CDN) has been considered by MNOs as an efficient delivery method for
popular content, such as blockbuster movies. The main purpose of having their own CDNs is to reduce
operational costs while providing good support to their core businesses. The deployment of CDN based on
big data technology is the most effective.
6.12.1 Big-data-driven cache server deployment in content delivery network
It is important to locate distributed cache servers in the CDN as close to the end user as possible in order to
shorten response time and also reduce delivery costs, e.g., a distributed cache server working together with
a central cache server in a hierarchical CDN. However, the cache access rate on the distributed cache server
might be lower than that of the central cache server. Sometimes the distributed cache server even needs
user data traffic to traverse the associated central cache server through the link in the event of improper
placement. Therefore, it is vital to choose the optimum location for the cache servers in the hierarchical CDN.
Beside the cost of storage and streaming equipment, the features and load of traffic in a given area are among
the important factors that determine the optimal placement of a cache server. After collecting data relating
to all relevant factors in the coverage area over a long period of time, big-data analysis can be used as a
feasible method in data analytics to help the MNOs deploy cache servers in the network.
The analytics capabilities are built into the hierarchical CDN by utilizing a collective intelligence data
architecture. Each cache server has a monitor agent to collect log information. This monitor agent sends log
status information to the data analytics function block, which then determines when and what content to
outsource, as well as placement of the replicas.
Therefore, the data analytics function needs to analyse the data related to both content and users to
accurately determine or predict content popularity.
6.12.2 Big-data-driven content scheduling in content delivery network
Content of high popularity is more likely to be placed on cache servers in RAN to improve the cache access
ratio popularity, while content of low popularity can be placed on cache servers in an edge provider. The
content popularity analysis depends on not only the content itself, but also users. Moreover, user mobility
may cause the content in the cache to change frequently, resulting in inefficiency in content caching.
Therefore, the data analytics function needs to analyse the data related to both content and users in order
to accurately determine or predict content popularity.
Overall, content popularity analysis and content scheduling require collection of the following data:
1) user data: the location and access pattern of users, which can be collected from the interface
between SGSN and GGSN;
2) content data: crawling from the Google or public website.
6.12.3 Content transmission control to end user of content delivery network in big-data-driven
networking
Due to physical space storage, cache nodes in a mobile network are limited, while cache nodes in a fixed
network specially deployed in the data centre are almost infinite. So, cache nodes in a mobile network can
only store limited content. There is a requirement to transmit data among cache nodes. Figure 6-11 depicts
content transmission control to the CDN end user in bDDN. The process is as follows:
1) a mobile user watches a video from the cache node deployed in the fixed network;
2) other mobile users request the same content;
3) the big-data plane, sensing the request, transfers data from cache node deployed in fixed network
to the cache node deployed in the mobile network that is nearest to these mobile users;
4) mobile users get the content from the cache node deployed in the mobile network;
Network and infrastructure 239