|© 2016 Black Swan Telecom Journal||•||protecting and growing a robust communications business||• a service of|
|Email a colleague|
Big Data? Forgive me for being cynical, but these days I’m suspicious of any noun with the “Big” adjective in front of it.
“Big” things often lead to big disappointments: Big Banks, Big Government, Big Debt, “Too Big to Fail” automobile companies, and “Big Transformation” projects that cost a ton of money but fail to make a telecom’s operating environment simpler or less costly to maintain.
If it was my choice, I’d substitute “Big Data” with something a little more descriptive like “Real-Time Data” or “Fine-Grained Data”? In analytics terms, those are the chief benefits you get from these cheaper and faster boxes: the chance to act on data quicker and view it in far greater detail.
Now while Big Data enables many new capabilities, its value is greatly enhanced when you can combine it with “Big Thinking”.
Big Thinking is what Benedict Enweani and his UK-based software firm Ontology are all about. And we’re pleased that Benedict has agreed to be interviewed in Black Swan.
First off, I must warn you to not be intimidated by that obscure word, “ontology”. Not to worry. Benedict does a great job of explaining his firm’s technology here. The Ontology guys are brainy, down-to-earth people, not philosophy professors tossing $10 words around to put you to sleep.
I’m not going to steal Benedict’s thunder, but I will say that the analogy to Sherlock Holmes style thinking is very apropos. You can apply deductive reasoning to investigate anything from mysterious murders to complex telecom systems.
As opposed to attacking integration issues with an army of Scotland Yard investigators though, Ontology uses a clever, repeatable method to orchestrate solutions at relatively low cost. My hunch is you can apply this stuff to many business assurance problems.
|Benedict, it would be great if you explained the origins of your firm.|
Sure, Dan. So, the genesis of Ontology, myself and my co-founder, Leo Zancani worked together at Orchestream, a mid-90s startup that focused on IP, VPNs, and internet services provisioning. Orchestream was acquired in 2003 by Metasolv, then soon after by Oracle.
And the thing we noticed in each and every deployment was it would take 8 out of 10 days to align data coming out of billing, CRM, and inventory systems so that we could just create the IP VPNs for the customers and put them on to the network.
This data alignment issue wasn‘t simply an Orchestream problem. In fact, every single deployment a service provider makes raised this issue. And the problem stretches from the Tier 1 giants to very small operators. BT, for instance, has about 3,000 different OSS systems for their core operations and even small service providers and ISPs have 10 to 20 different systems.
If you have a large array of clean systems, traditional integration technologies can actually bring data together quite nicely. Likewise, if you only have a few systems and the data is dirty, you can manually clean up those systems and get a joined up view of your customers, bills, and equipment at relatively little cost.
But here’s the point: whether you have 10 systems with very dirty data or 100 systems with moderately dirty data, there is not a single technical solution you can turn to that will bring those data together because the many misalignments are so hard to pin down.
And in many cases, the inventory and CRM suppliers are more than happy to take on these data cleanup projects. Some get rich from that type of work.
|Wouldn‘t it be nice if telecoms were like Amazon.com where it is like one-to-one relationship between a book sold and something in physical inventory? In telecom, things are never that simple.|
Exactly, we’re now helping Level 3 Communications to align their inventory and CRM. At face value, it sounds rather straightforward, but the problem is very complicated because there are something like 12 different systems where inventory data is stored.
The other complication we have in the telecom industry is that different internal consumers of data look at it in entirely different ways. In service provisioning, it’s about configuring ports on a Cisco router. In a Siebel CRM, you focus on a catalog of products with certain customer-restricted parameters. In a billing system, you worry about a line item associated with a particular customer order. In a performance management system, you’re collecting events and matching that with a service model and IP addresses.
|So how do you begin to address this immense complexity? How do you decide where to even start?|
Well, suppose we wanted somebody with a fresh perspective to come in and solve the problem. What if we went to the owner of a Mom and Pop shoe store and said, “Here’s our network operations center. It’s a mess and I want you to go fix it.”
Five days later, the guy from the Mom and Pop store returns and says, “Yeah, I fixed it.” And you stand there incredulous and ask him what he did and he says:
“I started with the network and saw where the actual services were and wrote those down. Then I went to the CRM system to look at the ”adds“ and ”deletes“ to determine what the network configuration should be. Then I noticed how you had forgotten to disconnect a bunch of circuits, so I freed those up. Finally, I updated the billing records to match the CRM.”
OK, what’s the point of my story? It’s to say that even the toughest integration problems can be solved in step by step fashion. Aligning systems is like solving a jigsaw puzzle where each piece you connect gives you a hint as to what to look for next. You arrive at answers by conducting a series of searches from one system to the next, checking the inconsistencies and reconciling views.
|But aren‘t you relying on an imprecise method of discovering where the errors lie? In a traditional systems integration approach, you’re examining the systems as they were originally designed and conceived. Isn’t that a more logical approach?|
Traditional integration may be the more logical, but it is not the most efficient path -- as the history of telecom integration proves. Integration is often an enormously expensive and risky business.
Consider this: the world’s biggest misaligned data system is not at BT, Verizon, or China Telecom. The world’s biggest misaligned data system is the internet itself. And yet the internet is very effective. People will often examine seven or eight web sites before they decide to buy something. The internet has become a key ordering channel for airlines, hotels, and many other industries, and yet nobody requires a fully integrated and joined view of the internet to get useful work done or to find answers.
The same goes for the systems at telecoms large and small. The point is to deliver as much alignment as is required by the mission, not to align all systems and spend a fortune doing so.
|What are some of the clues you use to connect the dots. How do you how do you achieve accuracy with your system?|
Well the first place to look is field names and id numbers. You try to match these across multiple systems and see what you find. An IP address is a unique identifier within a network. And wherever you have an IP address, it implies the existence of an interface. Likewise, an IP address implies the existence of a router or switch. If an IP address can‘t be associated with a network element, that breaks the “ontology” or rules of existence for things and therefore indicates an error in the data.
Along the way, we can correct a slightly wrong identifier. The whole point is to present the matches the system finds in a way that makes it easy for the network or system expert to make the final classification decisions.
In the area of network reconciliation, we can identify circuits that are likely to be stranded assets. We may not be able to find a match in the network side, but we can investigate services being leased with no associated bill to a customer.
|Other companies, such as Subex, are out there offering network inventory reconciliation solutions to identify stranded assets. How would you distinguish your approach from theirs?|
I know about TrueSource and it’s a fundamentally different methodology. Their starting point involves building a very extensive model of the underlying systems, and then pulling data from those systems so you can look for inconsistencies
Rather than start with a model, Ontology’s approach is to recognize the structure that is implicit in the data systems that people already have. Instead of pulling data from existing systems into another database, we pull it into a graph.
|When you say “graph” I think of a probability graph with a normal distribution curves. Is that what you mean?|
No, I’m using graph in the sense of a map, a fully connected node and edge structure that stores relationships: one entity connects to another entity through this particular relationship. And you can scan the nodes and examine all the relationships, so what you see is a kind of spider web called a “graph”.
What we are really storing is metadata — tags, pointers, and relationships among various databases and objects.
And as we examine the patterns in that graph, we look for consistencies, inconsistencies and when we find them, we classify things. In short, our approach deduces the structure of network and service delivery systems without having to know very much about the kind of data we’re looking at. And using this method we are able to deliver operational value in six to eight weeks, and the reason we can do that is we do not do a big integration at the beginning of the project.
Actually, Leo and I first got the idea for this approach from the intelligence community who pore through massive data sets trying to find terrorist cells. In that scenario, you have no control over where and when the data will change. Investigative work like that is a great example of teasing out the details of underlying systems using a centralized search method.
|Benedict, this is highly interesting stuff. It would be great if you could recap your perspective and value proposition.|
Well, Dan, it’s no secret that the telecom business is getting more complex. New services and new network technology are expanding. In addition, many new partners are entering the scene, be they: virtual network operators, cloud providers, IT outsource partners, over-the-top content providers -- even managed service providers.
But no matter how complex back office systems and partner relationships get, a telecom’s ability to squeeze out greater profits and efficiency depends on knowing a lot about customer behavior and network utilization that only a finer grain level of inter-system analytics can deliver.
So there’s a tremendous need to join data across multiple misaligned systems for business intelligence , optimization, and assurance purposes. But given the nature of today’s business, you need to do that at very low cost. And that tends to rule out those expensive and risky full-scale integration projects that were the norm in the past.
We tackle the problem a different way. We use orchestration and discovery techniques to align systems because they don‘t require you to change out or remodel your existing infrastructure.
What more can I say? Fortunately we became profitable last year and doubled our revenue the last two. So I encourage your readers to check us out.
Copyright 2012 Black Swan Telecom Journal