![]() |
© 2022 Black Swan Telecom Journal | • | protecting and growing a robust communications business | • a service of | ![]() |
![]() |
Email a colleague |
April 2014
The geologists who work at big energy firms like Exxon and BP know with great certainty where huge reservoirs of natural gas and oil lie 5 miles below ground.
But it wasn‘t until advanced technologies came along -- like deep water drilling and fracking -- that these giant firms could economically bring that crude energy to the surface and put it to good use powering the planet.
Mobile telecoms face a similar challenge. Their time- and location-sensitive big data can profitably be sold to marketers in other industries. Solution firms like Zettics and comScore even offer a turnkey way to analyze and enrich that data so telcos don‘t even have to touch it.
But a key bottleneck still exists: finding a way to access and manage these huge data volumes effectively. How do you economically pump the data in, store it, and filter it so it’s easier to work with?
Well, one company who has specialized in massaging telco data since 1999 is Sweden-based DigitalRoute. The firm’s claim to fame is in BSS where it OEMs its mediation solution to billing heavy-weights like Netcracker and Comverse.
But now, the great excitement around mobile data monetizing has brought DigitalRoute full-square into the OSS side too. So here to explain this trend and the many new data management challenges he’s seeing in the market is Thomas Vasen, the firm’s Product Management and Marketing VP.
Dan Baker: Thomas, it would be great if you could explain a little bit about DigitalRoute’s history in the mediation business. |
Thomas Vasen: Sure, Dan. We were actually quite fortunate. In the late 90s when DigitalRoute was formed, we could ride the wave of commodity hardware and blade servers emerging back then. And we’ve always designed our software to do end-to-end in-memory processing so that nothing touches disk while we process it. We take whole files in at once to give us a performance advantage, and we feel that’s still an important edge for us. We built one of the first truly convergent systems and can today handle batch and real-time data in a hybrid way.
Altogether we have over 300 active customers, most of whom have come to us through our OEM partners, though we serve some 60 operators direct.
What is it about your mediation approach that these OEMs like? Are you low cost? What’s the secret sauce? |
It is more complicated than people guess. We have focused on mediation and it requires expertise to do it correctly and efficiently. Over the years we’ve added more power and speed as you’d expect, but one point we try to emphasize is being partner-friendly.
To start with, the software is fully re-brand-able. Also, our partners build their own agents to take our technology and make it compatible with their proprietary solutions. We have also built one-step-upgrade types of functionality where you can correct, upgrade, and even easily roll back features. And our APIs are useful for embedding the software in a larger system. The customer doesn‘t even recognize that mediation from us is a separate module.
I understand an OEM you’ve recently attracted is SAP. What’s that all about? |
Yes, we are the data access point and collection engine for SAP who takes telco data and sells it as intelligence to advertising firms and marketing agents for example. The idea is to take raw geo-location and radio access network data and feed it enriched with demographics information from CRM, but anonymized of personal identifiers to SAP’s HANA databases. There they analyze it for marketing purposes and detect interesting behavioral patterns. Their service is called SAP Consumer Insight 365; it really unlocks and monetizes the network data.
Now the typical marketing agency who consumes this data might serve the leading department store in Sweden, as an example. And that retailer wants to know what kind of people walk by their stores but never come inside. That intelligence is of great value to them because if they cross-correlate demographics with location data, they can market to them much more effectively.
Now SAP is an IT giant with huge capabilities. So why can‘t they do the data gathering and filtering themselves? |
Basically they can‘t speak telco data at a deep level. Telco has very specific data formats, different aggregation rules, especially when you get into the radio access network world. It tricky to extract that. SAP and other IT vendors lack the capacity to process it accurate and efficiently.
What we do is intelligently merge telco data streams and put everything together into something manageable. This is why SAP is working with us.
What kinds of data do you capture? |
Well, our claim to fame, of course, is CDR processing for BSS applications, but DPI is an important new data stream we work with today. We don‘t capture DPI directly: we take in the records reported by the DPI engines or probes of Ericsson, Sandvine, Allot, Tek, Procera, and others. We then cross reference to identify what kind of service the customer has or the URLs being served.
This DPI data is invaluable in measuring customer experience. It’s not just about knowing what customers are doing, but also looking at dropped calls and sessions and figuring out what the customer did before and after the drop.
This data is all available but most mobile operators are not using it right now because the volumes are so massive, and they haven‘t learned how to manage it.
Why is that a problem? I mean, commodity hardware is here and standards like Hadoop are making access to large data sets more affordable. |
Hadoop’s vision is great, but the idea of plucking out data streams and saving them forever is not really working in the world of truly massive telco data.
We think the answer is to be real clever about the questions you ask of the data.
Most operators know exactly what they need to do with their network data. It’s pretty structured. The challenge is to filter it down to a reasonable size so it can be easily digested and worked with. Accessing the raw data alone doesn‘t work well because you are dealing with insane data volumes.
You need to prep and strategically look at your data. And it’s hard because in telecom the data off the switch is usually in binary form, not ASCII. So you can‘t just dump it into Hadoop and expect magic to happen. Data has to be converted to be able to correlate and unlock its value.
To give you an idea: a typical tier 2 operator in Europe takes in about 200 million records a day on the BSS side. And a tier 1 operator in the U.S. might be dealing with 6 or 7 billion mediation records a day for BSS. But that’s peanuts compared to the OSS side: an LTE network for a U.S. tier one is 2 million records a second!
And yet, DigitalRoute is able to manage that volume on a small set of standard hardware blades.
OK, we’ve talked about big data monetization, but what about the BSS side of the house? Any action happening there? |
As you know, Dan, the billing paradigm has shifted in a big way from minutes to bytes. And all the BSS logic around charging based on distance, area codes, time of day, etc. is not needed anymore.
But where there’s a lot of action in understanding the data usage -- counting billions of small records.
We call this capability “service control”, which starts with identifying the data records and routing them to a bucket that either counts subscriber usage or content provider usage. And our system allows you to create these buckets dynamically with an unlimited hierarchy.
Speaking of content, that’s a huge topic of interest. In particular, there’s a big tussle over exactly who is going to pay for streaming digital video. |
Yes, and a lot of early action there is happening in the U.S. market where some recent court rulings have relaxed the net neutrality laws somewhat.
And to me, these new regulations introduce a little common sense and give the service providers a little relief. The problem is that video subscription services like Hulu and Netflix are consuming an enormous amount of bandwidth across high speed Internet modems. Sandvine says that Netflix alone is eating up 30% or more of the North American data traffic.
So one remedy that’s emerging -- and AT&T was the first to announce it publicity — is to enable a reverse charging model similar to 1-800 and toll-free numbers. The idea is that the company who owns the content pays for the delivery charges.
And this is really the way it should be. After all, when you buy a book from Amazon, you also pay for the postage to have that book delivered to your home.
This trend is also playing into the cable industry too. Very recently Netflix and Comcast came to an agreement where Netflix agrees to pay an undisclosed amount for delivering its content to Comcast subscribers.
So anyway you slice it: either tracking usage or the number of times a user passes a retail store, sounds like there’s plenty of opportunity for DigitalRoute and its OEMs to stay busy. |
We certainly think so, Dan.
Copyright 2014 Black Swan Telecom Journal