Cox communications is currently overhauling network infrastructure to support gig-speed and full duplex bandwidth. This capital-intensive project needs a strategic plan providing prioritization using long term forecast accounting risk and macro-economic factors. Further, adding business constraints, budgetary restrictions, and other operational limitations we produce an optimal actionable plan.
The key business challenges with the current process of producing a prioritized plan are scaling, repeat-ability, and trace-ability. Today, there are few thousand nodes within the network, but within a few years, we will scale to over few hundred thousand nodes to meet network and customer growth. This increase will make it very difficult to continue with the current process of building a strategic network plan. Data preparation relies on a manual effort of extracting data from multiple sources to create a factual view of bandwidth consumption. Business rules are currently applied in spreadsheets using macros.
Automation of this process has opportunities on several fronts, notably, providing consistency, repeat-ability, and modernization with the use of data science algorithms on enterprise big data platform. Business requirements can be more transparent, and therefore configurable, allowing users to run multiple what-if scenarios faster to assist decision process.
A comprehensive end-to-end solution using modern Big Data platform will produce faster and repeatable results:
a. Real-time input data processing combining multiple network telemetry data (Hive, spark on HDP platform)
b. Forecast using an advanced statistical programming language utilizing CAGR and regression (Spark R on HDP)
c. Enterprise rule engine that can process declarative rule set (Jboss drools engine)
d. Specialized data visualization dashboards to drill-down and assist detailed analysis (Tableau connected to Hive)
Value additions by above solution:
a. Enriching bandwidth consumption data with external data from macro-economic factors, demographics, etc.
b. Augmenting forecasting process using proven time-series and ensemble based techniques
c. Apply artificial intelligence algorithms that factor cost of process, customer lifetime value, and competition to provide a mathematically optimal prescriptive plan
d. Easy to use workflow using rich user interface