The Evolution of Big Data & Analytics
Jim Hodges, Chief Analyst – Cloud and Security, Heavy Reading
I started covering analytics in a telecom context more than seven years ago when subscriber data management (SDM) was still an emerging technology. When conducting vendor and network operator interviews in April 2009 for the report entitled "Subscriber Data Management and the Era of Analytics," it became clear to me that analytics coupled with SDM could play a significant role in the delivery of highly personalized services.
Even then, the model had considerable value, despite the fact that it really only scratched the surface, since the big data component utilized a common back-end database to store only HLR/HSS, AAA and EIR data.
However, this SDM architecture is still relevant, in that it is still very much the same model utilized today; the only difference is that the SDM database referred to as User Data Repository (UDR) has given way to a massive, often Hadoop-powered database that can store data for any network node that impacts on customer experience. And this ability to holistically analyze any network data in real time that affects the customer experience is precisely one of the factors why I believe big data and analytics (BDA) will become a key network element in the very near future.
Of course, there are other factors to study as well. To do this, let's take a look at what has changed over the course of this seven-year span. It's actually a little bit frightening how much change has taken place over this period. The three major developments we see are as follows:
The rise of the mobile Internet and the Internet of Things
It's fair to say that in the past seven years, the mobile Internet has moved from fringe to mainstream service. Today 4G penetration levels and access to low-cost smartphones are really driving mobile Internet adoption, and also changing the service model. By that I mean mobile enterprise services are now a major market consideration, especially when augmented with social media or e-commerce capabilities. The end result for operators and enterprises alike is that they now have much greater business innovation opportunities to monetize the data. But to accomplish this requires that they put in place a complete BDA framework, to ensure that any promotions are contextually rich enough to meet the dynamic needs of the targeted users.
The other related consideration here is the Internet of Things (IoT), which will also reshape demand for storage and extraction of big data. While the current market focus for IoT seems to be initially focused on simply managing the data processing requests that billions of devices will generate, the long game is to leverage analytics to upsell targeted, personalized services. In both scenarios, however, BDA plays a critical role for ensuring that enterprises achieve the aggressive performance levels they need to serve their customers.
The impact of the virtualized cloud
Some four years into the network functions virtualization (NFV) journey, there is little debate that the application delivery cycle will be changed forever. But running telco applications as cloud VNFs also means that new approaches for concurrently processing mass data requests associated with orchestrating VNF lifecycles and failure recovery will also mandate new approaches for managing these requests to ensure ultra-low latency to match the high-speed performance of the RAN. This is another reason why a new breed of BDA systems that support a massive data processing engine and standard SQL or search-like interfaces and real-time data processing engine are vital.
The intersection of security and analytics
Just as BDA tools and techniques requirements are dramatically altered via a migration from the network domain to the cloud domain, so too are security tools. And in a short period, analytics has been identified as a crucial tool to help enforce real-time security policies. This will not change, and in fact looking forward, we see additional linkages as analytics tools become even more powerful and move to a predictive model that will allow security policies to be invoked based on pattern recognition and dynamic profile creation.
This blog is sponsored by Huawei.
— Jim Hodges, Senior Analyst, Heavy Reading