Past Entry in Zippy's Telecom Blog
October 27, 2009
A Talk with Håkan Millroth, CEO of Tail-f Systems, on High Availability, Open Source, and the New World of On-Device Configuration Management Software
Tail-f Systems provides on-device configuration management software and tools so that networking equipment providers can build carrier-grade network management systems. Their advanced software can be found in the realms of both Tier-1 equipment providers and next-generation startups. You can even find Tail-f System technology in enterprise IT environments. Tail-f’s software’s success results from its design philosophy: be scalable, be secure, be continuously available, be able to meet the needs of demanding environments.
One can divide the Tail-f Systems product portfolio into two major areas governed by their two major product lines, ConfD and ConfM.
First, their ConfD software solution helps equipment vendors imbue their equipment with on-device configuration management. ConfD is said to be the only data model driven toolkit allowing all critical northbound interfaces (NETCONF, CLI, SNMP and Web UI) to be rendered from a single data model. ConfD is so flexible that it can be used either to extend an existing management system or to implement a new configuration management system from scratch.
Both ConfD and another Tail-f product, Instant NETCONF Agent (a portable tool used to add a northbound NETCONF interface to an existing device), are designed for embedded systems.
ConfM, on the other hand, is designed to extend the functionality of EMS/NMS products by enabling a southbound NETCONF interface to be added to an existing element management system.
The reader will notice that NETCONF keeps appearing in these discussions. The Internet Engineering Taskforce’s (IETF’s) NETCONF protocol handles network device configuration, allowing devices to expose an XML-based API to the network operator. It can be used to send and receive full or partial configuration data sets. The NETCONF interface provides an implementation of the NETCONF protocol, as well as a complete mapping from NETCONF operations to configuration database operations for either the integrated data store or a data store of the developer’s choice. Tail-f contributes to the development of relevant industry standards in our technology area, and they actively participate in the Working Group for the NETCONF standard within the IETF (RFCs 4741 and 4742).
I recently had a wide-ranging discussion of these matters with Håkan Millroth, CEO Tail-f Systems. Before co-founding the company, Millroth held an executive management position at Interpeak, an embedded software vendor that was acquired by Wind River in 2005. Previously, he was a co-founder and CTO of Bluetail, a software company acquired by Alteon Websystems in 2000. Mr. Millroth’s career also includes several management roles at Ericsson and Nortel Networks. Mr. Millroth began his career as a Professor of Computer Science at Uppsala University, where he conducted experimental research in distributed systems.
RG: From where did Tail-f Systems get its unique name? I suspect it’s from Unix.
HM: Yes, the Tail command in Unix is used to look at the tail of a file, and the “-f” is for monitoring. We’re into network management and the most basic form of management you can do in Unix is to watch a log as data is written into it. As you might have guessed, we have many Unix veterans in our company. It’s a name that appeals to geeks.
RG: “Commoditization pressure” from the bottom up is hitting the middleware layer with the OpenSAF project. Although the project is in its early stages, many say that it’s sure to present some interesting challenges concerning how to position and maintain margins on proprietary implementations. I guess the present economy is helping open source considerably.
HM: Basically I would say that most of our customers use some kind of HA middleware layer. There are three major vendors in this area: Enea, GoAhead Software and OpenClovis. Of these, OpenClovis is the best known company in this area that incorporates an open source approach. Clearly, during 2009 we have seen an uptick in interest for open source software. At Tail-f, our products operate at a layer “above” that. Our ambition is to make it possible for pre-integrated solutions to exist incorporating our ConfD product and those from these other companies, so that basically the customer can choose our management system independently of whatever HA solution they happen to be using.
Having said that, we are part of the OpenSAF project. We also support the Service Availability (SA) Forum’s Information Model Management (IMM) service. Tail-f uses this standardized service as an add-on module to our ConfD product so that carrier-grade network management solutions can be deployed that can utilize such useful interfaces as CLI, NETCONF, SNMP and Web UI. The IMM is part of the SA Forum’s huge Application Interface Specification (AIS) for high-availability middleware that enables developers to write HA applications via a set of standard interfaces so that the apps will port across various vendors’ AIS implementations. AIS consists of a lot of things, but one of them that is of particular interest to us is IMM. It basically comprises what is necessary to manage the rest of an SA Forum-compliant middleware system. We applied some of our know-how in system management and contributed that to the IMM specification.
So we’re definitely part of both the SA Forum and the OpenSAF project, but that for us doesn’t rule out our working with the commercial vendors, some of which have SA Forum-complaint systems, others are not taking the SA Forum route, and still others are developing their own “mix” of the two approaches. In fact I would say that today in the HA industry most vendors have a mix of SA Forum-compliant components and then they add value to that with some of their own secret sauce that might not yet be standardized.
So in a sense we are represented everywhere. At least in the real telco equipment market involving Network Equipment Providers (NEPs) we are pretty much becoming the dominant commercial vendor and therefore it is also important for our customers that if they decide to use our ConfD product, they can be assured that it works with all kinds of HA systems.
RG: With hardware, OS and HA/Middleware vendors increasingly switching from proprietary point-solutions and working towards pre-integrated solutions for customer Network Equipment Providers (NEPs), who will take the integrator role? IBM? One the one hand everyone wants flexibility, but on the other hand, there is the convenience afforded a pre-integrated solution.
HM: We do see a quite clear split in our customer base: Some vendors are very much looking for someone like IBM to take the lead and be responsible for packaging together the hardware, operating system, HA middle and the management system and thus take on the role of an integrator. Other companies, however, are more into picking best-of-breed solutions and doing the integration themselves.
RG: I guess like everything else, it depends on the corporate culture.
HM: Yes, when it comes to the embedded software in the networking boxes, it’s also a matter of the company’s “culture” in terms of whether it’s an organization that has been operating for 30 years and has a very strong internal key architecture and culture. Such decisions are really quite people-dependent and corporate culture-dependent. Interestingly, the exact same phenomenon can be found at the network level. Some service providers still are picking best-of-breed boxes, others are looking to IBM to integrate and others would like to see Siemens be the integrator, to integrate solutions incorporating other vendors’ boxes, and deliver a whole network to customers, rather than just a bunch of boxes. Once again, different cultures can be found in different companies.
In general the trend is toward more pre-integrated solutions, both within the network equipment area and across the networks themselves. That’s also clearly a trend we’ve seen in big IT departments over the past 10 years.
Applications are starting to increasingly emerge from “the cloud,” so the question arises as to who is exactly the integrator. Is it the company that builds the cloud and manages it? It’s an interesting area.
RG: Virtualization features running on network devices are sure to take off into mid-range equipment in 2010. Where is HA in all this? Running on top of hypervisors or in a new layer?
HM: It’s difficult to say. In 2008 we didn’t see any customers using hypervisors, also called Virtual Machine Monitors (VMMs), which are computer software/hardware platform virtualization software enabling multiple operating systems to run on a host computer concurrently. Now, however, we’re starting to see some companies use them.
How that affects the HA layer or what’s happening higher than that, where our management software functions, is a completely open question at the moment. Hypervisors are a nice thing to have when you’re building cloud datacenters. In fact, the whole concept is based on virtualization. That is the reason that we see some of our equipment vendor customers having an interest in supporting virtualization in their devices. But I think it’s all very much in flux how things will end up. On the one hand, you have network virtualization, then you have host virtualization, and then you have storage system virtualization. It all comes back to a more fundamental question, which is how cloud computing in general will come up with a unified way to manage networking, storage and computing. That is what everyone is focused on right now. That focus became much sharper when Cisco announced their Unified Computing Initiative earlier in 2009. At that point, many companies started to think about what their strategy should be for unified management of networking, storage and computing. Thus, cloud computing is driving a lot of networking decisions, even in the vendor space.
Another major trend we see is that mobile data is finally taking off big time, which has a major impact on networks and therefore a major impact on our customers who are providing equipment to the network. So there is an intense focus on mobile backhaul, carrier Ethernet, video streaming, overlays on the network, and those sorts of things.
RG: Basically, Tail-f focuses on device configuration management on the one hand and things like NETCONF on the other.
HM: Yes. We have assembled an impressive customer list three-and-a-half years after shipping ConfD. NETCONF is a key component of our ConfD. The adoption of NETCONF by the equipment vendors is definitely occurring. In the case of all of the Tier 1 carriers and most of the smaller companies that we talk to, it’s more or less a given that they will have a NETCONF capability in their next-gen boxes. I see NETCONF as an enabling technology for the next step, which is how you best use the NETCONF-enabled network, and then you can start looking up toward the network management layer and see how you can best utilize it there too. That’s where we’re focused.
RG: Thanks for the interview. We’ll be waiting to see what Tail-f comes up with next.