WTRS Executive Interview

Interview with Dr. Håkan Millroth, CEO of Tail-f Systems.

September 10, 2008

George West: Could you tell me a little bit about your background and how you came to Tail-f Systems?

Håkan Millroth: My background is actually in academic research. I started out doing research on distributed computing back in the late 80s and early 90s in Sweden. At Uppsalla University I had a research group with four or five graduate students. What happened was that we had a research collaboration with Ericsson Research, and basically Ericsson asked me if I could move over to Ericsson and start up. At that time Ericsson realized that they were a software company. They had always been a phone switching company. Then they came to the conclusion that the value of what we do is software.

They asked me to start a software architecture center at Ericsson to try to assemble the best practices of software architecture with the ultimate goal of building that into platforms that they could use across the organization. I took two of my newly-baked PhDs with me to Ericsson and looked around the organization for a year for some young, ambitious and really good architects.

I assembled a group of 8 - 9 really top notch people and our approach was that we data-mined the most experienced system designers and architects across Ericsson so that we ourselves got an understanding of how do you really build 5-9's, traditional Telco, systems. Ericsson had an enormous amount of knowledge distributed over the organization.

Then what we did was we went out to a couple of new product organizations when they were just about to launch new development efforts and said, "Let us drop three guys here to be the architects of your team. We will stay here for six months to build the Christmas tree and then we will leave a guy behind to stay with you for a year to hang up everything on the Christmas tree so you get a great architecture." We did that in three different areas of Ericsson and that worked perfectly well.

Then in 1998 when this whole Internet started happening we realized that this is exactly what the Internet is. During the process of building big server farms was when people started using the Internet in transactions involving money. And Ericsson said "We are not at all interested, that is completely out of our area." So I sort of banged my head for six months and we decided to start our own company using that business model.

We started a company back in late 1998, called Bluetail, where we built a similar kind of platform. We used our know-how from Ericsson on how to build a distributed cluster platform. Then we partnered with a US company called Alteon WebSystems that was doing load balancers and Layer 4-7 switches to supply them with that platform.

Quite soon after our licensing agreement with them, they acquired our company. That was in 2000 and almost immediately after that Nortel acquired Alteon so myself and my whole team ended up at Nortel, another telecomm that didn't understand how to do the Internet.

I left Nortel in 2003; most of the other guys were staying until 2004. And then we decided to focus on the specific part of network management of that whole platform. We had already tried building a whole platform but that was too ambitious. It was a great idea and it actually paid off in the late 90's but in today's environment you shouldn't be that ambitious you should focus on a subsystem part of the problem. So we focused on the management part of it and started Tail-f Systems.

One of the reasons that we started Tail-f was that we realized that everyone is using Linux in new projects, everyone is using pretty much the same signaling stacks, if they are using router stacks there are two or three different versions of it. If they are using a SIP stack there are two or three vendors for open source

This means that the things that you manage from one project to another project are very similar, on similar hardware. That means that there is an opportunity to take out the whole network management piece within the boxes, which makes up 30% of the typical project and actually packetize that as a product and that solves 80% of the problem. Not 100% but 80%; 20% is always different because people have legacy software and what not.

Our idea was let's solve 80% of that problem in a product. Let's use our experience and understanding of how this stuff is actually built in practice to design really elaborate and good APIs so people can extend it to solve the remaining 20%. That was basically our business idea when we started.

George: How did that evolve into the NETCONF standard and into the products Tail-f Systems is shipping today?

Håkan: A little bit before we started Tail-f, NETCONF had made some real progress within the IETF. One of my cofounders, one of the engineers, Martin Bjorkland, he was involved, just because he was interested and knew a lot about this topic, he was engaged with that standardization work within the IETF. We realized that this was definitely a new thing in this area of embedded network management, that we should make it the cornerstone of our product because it allows us to open a lot of doors because it's a new thing.

Although from a functionality point of view NETCONF was only 25% of our product, it was a new thing that happened at exactly the right point in time. If we had first built our product and then NETCONF became a hot topic two years later, we would have been toast because it is very hard to build that in afterwards.

Lucky timing I would say; that NETCONF and when we were founding the company were more or less simultaneous.

George: What is important to understand about NETCONF the protocol?

Håkan: First of all, it is XML based. Before people started using NETCONF a lot of vendors, after the year 2000, were starting to use XML. The first company that actually provided an XML based management system was Unum. And then other people followed that and built proprietary XML based management interfaces for boxes.

So the industry realized that using XML for configuration was a good thing. That was common knowledge in the industry. NETCONF is XML-based and that's a must in today's network.

The other thing that is important to know about NETCONF is that you can do transactional changes across many network elements at the same time, which ties in nicely into service provisioning. When you provision a service you very often need to touch many boxes when you actually implement that service activation on the boxes.

These are the two main things that are important with NETCONF.

George: As part of the Tail-f product suite there are two functional areas. One is CONFD which runs on equipment and the other is the CONFM product that runs in the framework of an OSS/EMS system.

Håkan: CONFM runs as a plug in an OSS. It is essentially the configuration management subsystem for an OSS.

George: As a network provider what does this functionality allow me to do? What can I do when I have this distributed configuration management functionality?

Håkan: What it can do is tie together your equipment with your service provisioning mechanism.

Today service provisioning ends at the EMS layer. This means that somehow you or your equipment vendor must solve how you make things happen at the actual equipment. That requires some magic at the EMS layer or most often it never happens. It allows you to actually take advantage of the entire chain from service provisioning, service activation in real time, in IMS networks or other networks, and automate that process all the way down to configuration changes on the real equipment.

You can't do that today because that level of automation stops at the EMS layer.

George: How does NETCONF help you manage the complexity of multiple vendor equipment or multiple versions?

Håkan: Today you have the problem that as long as you stick with one vendor or even one kind of product from one vendor, you probably have a pretty good network management all the way down to the equipment. As soon as you bring in multiple vendors you can't manage across vendors in a unified way.

You have different ways of managing your CISCO equipment and your Ericsson equipment and your Alcatel Lucent equipment. There is standardization on the OSS layer but there is no standardization when it comes to configuration management from the OSS down to the equipment. It basically allows you to truly have multi-vendor networks. And that's a big advantage I would imagine for many carriers.

George: Are there any points that I missed or questions that I should have asked?

Håkan: Ultimately where I think that NETCONF as a standard, and the products that we build around NETCONF, will have the most value is in the OSS space. That is where you need built-in support for NETCONF. That will only happen when the equipment vendors have adopted NETCONF and that is happening big-time right now. Either now or very soon the OSS industry will pick up that new technology.

George: For Tail-f there is significant opportunity to work with the OSS vendors.

Håkan: I think so.

In the long term, we are well positioned since we are starting at the equipment side and if we haven't done that, we haven't created an opportunity for doing something game-changing on the OSS side as well.

George: Thank you.

More information about Tail-f Systems here...

This interview ran in our September 10, 2008 newsletter issue.