Blog Post

Anagran Launches: The Return of Larry Roberts

[qi:033] Dr. Larry Roberts, one of the key people that helped build ARPAnet, the precursor to the modern Internet, is taking a second shot at entrepreneurship with Anagran, a Redwood City, Calif.-based start-up that’s building an intelligent flow router designed to enhance the quality of packet-based networks.

The company has raised $28 million in funding from Advanced Technology Ventures, Arrowpath Venture Capital, Draper Fisher Jurvetson and Argon Venture Partners. (We wrote about the company back when it raised some early VC dollars.)

Roberts’ previous company, Caspian Networks, flamed out after blowing through over $300 million. Unlike the last time, Roberts argues that Anagran is benefitting from the progress made in chip technologies and leveraging powerful off-the-shelf technologies.

A flow is a single meaningful end-to-end activity over the network, and is defined by the IPv4 header 5-tuple of source and destination port, source and destination address, and protocol. Examples of flows would be a video download, a voice call, or an image transfer.

Anagran’s device is essentially the next generation of traffic management, a concept that Roberts has talked about in the past. As video, voice and other specialized traffic flows on IP networks, Roberts argues that a packet-based approach isn’t going to work. When a video is being streamed, any dropped packets can cause quality degradation. Voice-over-IP services also have similar issues. However, if the router is focused on an entire stream (or flow) of video or voice, it can manage them more intelligently, and take corrective measures. In addition, Anagran is also well-positioned for the IPv6 transition, since IPv6 has the same flow-type characteristics.

“This really is the router of the future,” he told EETimes. Others disagree.

“While there are certainly benefits to [Anagran’s] approach, I don’t necessarily see it as a superior alternative [to typical routers],” David Vorhaus, analyst with The Yankee Group told Telephony Online.

Despite the pedigree of the founder, the fact remains that getting into the router-market and squaring off against Cisco Systems (CSCO) and Juniper Networks (JNPR) is not an easy task. Several have failed, and others have had to embrace equally large players like Alcatel-Lucent (ALA) to earn a living. We will keep an eye on this one: it is, after all, Larry Roberts!

6 Responses to “Anagran Launches: The Return of Larry Roberts”

  1. Jon heil

    Certainly one must be able to limit customers traffic for non revenue generating traffic and p2p users should not make the provider over provision their networks. If one put in place a mechanism to limit the number of video and voice calls in high priority queues, this would be a welcome addition to any IP network. Its fine to let a network congest if people are using the network illegally for p2p stuff. Why should I bear the cost of another 10G interface?

  2. While no-one can predict what will happen to any one company it is both easy and safe to say that within 2 generations all IP routers will have strong dpi and flow management capabilities. Most already have first generation flow-like technologies designed for streaming (ldp for example) forwarding. It’s essentially the same idea.

    The Caspian platform was ahead of it’s time in many ways, although — again can’t argue about the spending of money and the making of companies. Good technology doesn’t necessarily make for great products.


  3. I’m definitely not sold on the concept. The trouble with this approach is that it expects that it can say which packets can be dropped and which packets can’t be dropped. Like any and all Quality of Service mechanisms in the network it is based on the premise that there will be times when the network needs to drop packets. However a well run access network and backbone will not ever need to face that situation, if the network upgrades fast enough. If the network doesn’t upgrade fast enough and it meets bottlenecks than these kind of technologies might alleviate the pain for a short while, but with traffic growing in most networks at a rate of 50% to a 100% per year, pretty soon after you’ve chosen to drop one packet, your dropping one more and than one more at an alarming rate. By the end of the year, not even real time voice and video are safe from dropped packets.

  4. While the founder’s background is a plus, I’m not sold on the concept. I’m not even sold that IPv6 will happen. How are we supposed to remember IP addresses? Its impossible – they are too long.