Former Time Warner Cable CTO LaJoie Provides Download on Big Data

After spending more than 25 years in the cable operator industry, former Time Warner Cable CTO Mike LaJoie has dealt with pretty much every network issue imaginable.

LaJoie, who also held the position of executive vice president at Time Warner Cable Inc. (NYSE: TWC) prior to retiring near the end of 2014, started out at Time Warner in 1994. (See TWC's LaJoie Leaves Parting Patent Gift.)

Among other engineering-related feats, he played key roles in the company's Full Service Network, which was a precursor to how networks function today, the first launch of broadband by a cable operator, and the rollouts of Time Warner Cable's network DVR, VoD and VoIP services.

LaJoie hasn't rested on his laurels after retiring. He is an advisory board member for SD-WAN startup Viptela as well as board member for Envivio Inc. (NASDAQ: ENVI). LaJoie is also chairman of the board of directors for Big Data analytics firm Guavus Inc. .

LaJoie was wearing his big data hat during a conversation with Telco Transformation.

Telco Transformation: Mike, to start with can you talk about the impact that big data is having on service providers' digital transformations, and the different ways that data is being collected and analyzed?

Mike LaJoie: There are a couple of areas with big data. One is the collection of data and then actually processing that stuff, enriching it and making sense out of it. Changing data into information and then finally coming up with information that you can take action on is what the big data transformation is all about.

So lots of folks say they can get big data. There are tons and tons of information being generated, but most of the companies that are out there that generate useful information from large data points are utilizing probes. You have to put hardware somewhere out on the edge and then sample that. That hardware sits there and gets into the stream and it samples that probe. That all gets aggregated and gets stuck in a big hardware drive somewhere and then you have this big lake of data. You go crunch that and enrich it by combining data from other sources. Some days or weeks later you get information that is a nice view out your rearview mirror. If it's really going to be transformational, you want to be able to look forward instead of backward.

The way that most people are doing big data analysis now is you get interesting information, but the problem is it takes so long to create the views into that data and you're looking at past events that have happened. It's very, very hard to translate that into "OK what do I need to change, what do I need to differently?" It's even harder to recognize when a bad event that happened in the past is happening again, or even better, is about to happen again. It still takes time to crunch all of the data and it takes time to do all of that analysis.

TT: So that's one way that big data is being collected. Can you give us another example?

ML: We [Guavus] try to utilize the information that comes out of the devices themselves. So garnering information from edge routers, from call centers, set-top boxes, from IVRs (interactive voice responses), from wireless phones or from the RAN (radio access network) itself. We do those devices instead of putting a bunch of probes out there. So if you want to go figure out more information about what is happening with people who are using their cellphone as a tethering device, you'll have to install a bunch of probes all over the network. You'll have to install hardware that's going to look for a particular anomaly. If you're able to tap directly into the RAN itself, or even better put a client on the cellphones that are generating this information, then you have a much shorter path getting to this data.

OK, now I know that tethering is unpredictable and it creates a load on my network that I can't accommodate, right? So what do I do about that? I have to be able to spot it when it's happening and I have to be able to change my network practices and my operating practices. I have to either let customers know that if they do this it's going to cost more money, or if I spot you doing it I'm going to have to throttle that particular stream or those streams for those people that are tethering and slow them down a little bit. Otherwise it swamps my network.

Being able to recognize that's happening is the real secret sauce of what Guavus does. Guavus can develop the signature and when the network is starting to promote a signature of a lot of tethering in a certain area, say a college or university, you can turn on the throttling protocols. You can't do all of that stuff in cable because the FCC will come get you, but you can still do it in wireless.

TT: So you've talked about how big data works for wireless service providers, what about cable operators?

ML: I think the big opportunities are in cable. By actually looking to several sources of data and comparing network performance, change management, terminal performance and call processing at the customer care center you can do the analysis across those sets of data and find out why you had a call spike Wednesday morning. Now we can look for concurrent anomalies and we can drew conclusions from them.

With big data, we can find out that the big spike Wednesday morning was due to a change management that was made Tuesday night or Wednesday morning. In the past it would take weeks to make that analysis and come up with that conclusion. Now it can take minutes or hours.

TT: So are we talking about big data on a micro level or macro level?

ML: Let's say in the Carolina region we have more phone calls in a particular period of time than other regions. Looking at that from a large-scale perspective you would ask why that happened? In real time you can further analyze where in it happened within that region. It's not the entire states of North and South Carolina. It was actually driven by a problem in Greenville, South Carolina. Greenville had a 30% spike in the Carolinas. Then you can dive into that real quickly to see what happened in Greenville. With more sensitively you can start to see that signature develop for what happened in Greenville or what is happening in Greenville and how to approach it.

With that better visibility and being able to identify those signatures, you can see that it was the change management Tuesday night that screwed this up, or we had a particular router fail in that area. With a router, if you are paying attention, if it's starting to drop packets, or if it's starting to get flaky or if you're getting a lot of flapping in a particular node in the network, you see that before the phone rings with someone saying there's a problem. If you look at that signature you can see that the router started failing hours or days before the phone started to ring. That kind of information is truly actionable and really allows you to get efficiencies.

TT: So what about Netflix Inc. (Nasdaq: NFLX), or any another over-the-top provider, are they tapping into big data as well?

ML: Yes, they certainly are. Anything that gives you visibility into uncharacteristic anomalies that impact network traffic or service performance, this kind of real-time analysis will help. Netflix is an interesting example. They build and operate their own stuff in a proprietary fashion. They've built a lot of this stuff into it, but a company like Netflix, even though they have scale, there's no way they're going to keep up with all of the money that is being spent on analytics. All of the money that Splunk Inc. and others are spending, they're not going to be able to keep up with it. They're going to need to take advantage of those tools.

TT: So we've heard a lot about what bg data can do for service providers, but what needs to be done on the backend to provision it?

MJ: That is part of the issue. You've got these tools, but how do you integrate them in? When you put the Guavus solution in place, it takes them several weeks to integrate it into the overall systems. They work with the operator to decide where they're going to pull all of this data from. Do you pull it from IPDR [Internet Protocol Detail Record], out of the CMTS, or do you pull it from the call records on their IVR? Or do you pull it from voice analytics of phone calls? Or am I going to pull it from Netcool alarms? When you go install the solution you work with the operator and you say: "These are the places we need to pull the data from together." Guavus takes that data normalizes it, enriches it and is able to process it in a real time fashion to provide actionable information.

TT: So what's next for big data?

ML: If we can take these same functionalities out to enterprises and have a set of solutions for a broader range of businesses and not just service providers, that's kind of the Holy Grail.

— Mike Robuck, editor, Telco Transformation