Mining data mountains for gold
Operators amass huge amounts of data from huge amounts of customers. Yet they don’t fully exploit this un-mined gold from millions of people. In the telecoms sector, the market value of big data analytics is growing - Heavy Reading predicted an increase of 26 percent year-on-year to hit US$3 billion in 2015.
What is big data? How do we define it specifically in relation to operators?
1) It uses processing technology such as intelligent storage, intelligent data mining, and intelligent analysis to extract commercially valuable information. From the huge amounts of scattered and fragmented data that operators possess, the resulting information is applied to industry-specific applications. Operators can use it to make more informed judgments about trends, perform precision marketing, and optimize how they do business.
2) It requires a sufficiently large amount of data – normally at the PB (petabyte) level. Macro-level patterns can only be analyzed from micro-level data in sufficiently large quantities and collected over a long enough time span.
3) Big data technology is fundamentally different to traditional data mining technology. They differ in terms of data types, processing mechanisms, and processing speeds.
4) Applying big data will increase the productivity of society and change peoples' lives with user-centric approaches.
Big data is driving transformation
More carriers use big data to drive business transformation like network evolution, Internetizing the user experience, and satisfying diverse, personalized user demands. They’re also using it to out-compete OTT and MVNO players, innovate services, and find new revenue streams.
A real-world example includes AT&T. The American giant predicted and prevented potential churners by collecting and analyzing data from different fields and taking preventive measures, increasing retention by 36 percent.
Equally, Vodafone used big data to optimize network KPIs, reduce complaint rates, and lower CAPEX for capacity expansion, while Telefonica applied big data to grow and monetize its digital services.
In terms of building big data capabilities, operators still have a long way to go. So, they're looking away from infrastructure and toward application-based operations. Key features are enabling business, building platforms, and establishing big data ecosystems, with the aim of maximizing the business value of big data.
Hi, I’m FusionInsight
Huawei's big data solution FusionInsight is the culmination of Huawei's ICT and big data knowledge coupled with its understanding of operators' pain points. These include finding ways to use big data to know what users want so they can deliver it and better manage demand. Another is competing in a saturated market with rivals who are winning new users and encroaching upon their customer base at a worrying rate. Operators also need to maximize the potential value of their digital services and find new business models and profit sources.
Huawei's big data solution utilizes layered architecture, the core of which is the FusionInsight platform and its two layers: data platform and data service.
The platform layer hosts cross-field data, and the service layer supports applications. It delivers the following technical solutions for operators: efficient storage and processing, an integrated data platform, real-time streaming technology, E2E scenario modeling, innovative data monetization, and unified data operations and management.
Basic infrastructure
Huawei's enterprise-class infrastructure for big data is unified, and centrally manages data assets from different subsystems. This resolves issues such as repeated data collection by traditional siloed IT systems, fragmented storage, and high construction costs. The solution offers a low-cost, high-performance solution that meets the challenges of massive data. In 2014, it turned a profit in the storage market in China, and in Q4 was the fastest growing solution in the global storage market.
Data platform layer
The data platform layer meets requirements for data convergence and real-time processing, and also efficient, low-cost data collection, conversion, storage, and processing.
Its functions are as follows:
- Scheduling cross-Hadoop cluster tasks on more than 100 large-scale clusters and unstructured files.
- Dynamic upgrading to double the performance of ETL processing.
- Accessing Hadoop components via Kerberos for data security.
- Improving performance by increasing upload efficiency in masses of small files, with 100,000-flow scheduling latency in less than 30 seconds.
- Providing API interfaces for flow configuration and execution for external services.
Data service layer
The data service layer includes cross-field, full-feature analysis such as in-depth service modeling, propensity forecasting, theme extraction, character profiling, relationship estimation, characteristic analysis, and automatic feature construction.
It also incorporates advanced streamlined algorithms from Huawei, and the modeling library from its Noah's Ark Lab. These functions provide powerful support for operators to build big data modeling capabilities.
The data service layer is rich with outstanding features:
- Cross-field analysis, BOM full-feature analysis and modeling, support for 1,000+ user characteristics, which will soon scale up to 10,000+.
- HiGraph modeling algorithm developed by Huawei, which is three to five times faster than MLlib and has an average AUC accuracy of 75 percent.
- Openness and seamless connection in R programming language.
- Automated modeling, character selection, algorithm optimization, and derivative closing.
- Reduced modeling times from 2 hours to 10 minutes.
Application layer
The application layer of the FusionInsight platform can construct service-oriented applications decoupled from the data layer. These include precision marketing, user experience management, higher network efficiency, data monetization, and other external monetization applications.
The application layer offers the following features:
- Full cross-layer security management from L2 data collection and conversion to L4 external cooperation.
- Full privacy protection, authenticated access based on login identity, and user data management based on user groups and irreversible encryption.
- Various ways to empower data openness.
- Easy to use third-party development environment that supports the rapid, low-cost development of third-party applications.
Control center
The big data platform control center manages data assets, data security, work scheduling, and operations so operators can unify how they use big data.
Interlinked big data service capabilities
Huawei delivers end-to-end big data solutions. It has all-round strengths in consulting, services, hardware and software, and integration, which helps operators to rapidly implement suitable big data applications to achieve business and operational transformation.
Huawei's R&D centers span the globe. Our services are interlinked and we provide end-to-end service solutions from initial scenario analysis to business modeling and training in modeling. We help operators build big data capabilities.
For Indonesian carrier PLDT, we delivered a solution comprising big data collection, convergence, and big data infrastructure, which cut hardware costs by 50 percent and boosted ETL capabilities by 30 percent.
For China Unicom Shanghai, we built a big data ecosystem to help create its monetization business model, generating 10 million yuan (US$1.54 million) in revenue for the operator in 2015.
Huawei bases its solution on fully understanding operators and how to best extract gold from data.
