RECAP: Moving Telcos to Real Time
Data and Analytics Forum: Moving Telcos To Real Time meeting recap by Telecom Council Staff
Timing is everything. True in comedy, in logistics, in love…and in telecom, as it turns out. This week the Telecom Council’s Big Data and Analytics Forum focused on the subject of Real-Time Data and Analytics at a meeting hosted by Dell. Telcos historically have run batch jobs on static databases overnight, because the systems and the data itself was just “too slow”.
This has meant that great ideas could not be put into practice, because the reaction times were just too slow. But by moving data and analytics to real-time, network operators open up opportunities to:
- create marketing/sales offers (for carrier’s services) within seconds of a customer’s need
- create relevant ads based on the location and recent activity of the subscriber
- react swiftly to security attacks, DDOS, viruses, botnets
- provide customers up-to-date, error-free usage stats in a dashboard
- react to changes in network utilization, system faults, demand spikes
- provide intelligence that can be used by a SDN to manage the network and virtual assets
- predict and forecast the above-mentioned opportunities and problems
The change to real-time data is neither impossible, nor easy. Legacy systems still make up the bulk of the network. So to get to real-time, the first solutions slot-in to specific roles, or work as overlays on top of existing OSS/BSS systems. In many cases, the data still goes to the older, slower system, but the data flow is forked, and one path of the data takes it to a real-time system capable of faster storage, and faster processing. The conventional features (ex: CDRs, a monthly bill) still come out of the old system, but the new real-time features come out of the modern overlay. Amdocs noted that conventional OSS/BSS vendors and database providers were not standing idle, but also have solutions for a real-time data world, and a roadmap to help carriers migrate from their legacy products to modern platforms.
In our meeting, the audience wanted to know what, specifically, was required to get to real-time. The panelists answered that it required faster database technologies, powerful datacenters, flowing data as opposed to static, and the creation of elaborate business solutions built on top of that hardware to meet a specific business need. No dealbreakers, just some hard work.
Dr. Uhlig, CTO of Adello, dug from his experience as a High Frequency Trader on Wall Street to explain how the take-aways from that industry apply now to marketing opportunities in telecom. Dr. Uhlig also explained a cyclic development process where models are built, simulated and tested, then deployed multiple times per day. Real-time for the models is a matter of hours. But once deployed, those models make snap decisions about what a customer might want, and real-time for that decision is a matter of milliseconds. The results are accumulated, added to the data, and new models are built, completing the cycle. At some point, data scientists in a cross-function DevOps team need to inform or “teach” the models…but most interesting is that machine learning or AI may be able to ease humans out of this process loop.
We also heard from a number of startups and vendors with solutions available to help the network operators make the shift to real time. Presentations included: Dell, ADello, VoltDB, Emagine, NetScout, Hitachi Systems, Argus Insights, Sandvine, Impetus Technologies, NetNumber, Morpho, Amdocs, Tupl, Cask Data, MemSQL, Kentik Technologies, IOVOX, M2Catalyst, Cinarra, Rocana, and Bottlenose.