Blog Post

Can Networking Be Made Cool Again?

Remember when networking was cool? When I started my career in 1997 as a wet-behind-the-ears intern in Yankee Group’s data communications practice, networking was the coolest thing around. Billions of dollars were flowing into the market to drive the emerging Gigabit Ethernet wave, build out the Internet core and jump-start the carriers (CAPs/IXCs/ CLEC/BLECs) that had been spawned by telecom deregulation the year before. Some, such as Juniper, Foundry, Extreme and F5, became well-known players in the space, while others, like Qtera, Xros and Sirocco, were gobbled up in billion-dollar-plus acquisitions, never to be heard from again. Far more were crushed in the collapse of the Internet and telecom bubble, leaving behind little more than memories of over-the-top launch parties.

Source: Dow Jones VentureSource

And as the chart above makes clear, networking and communications have fallen out of favor, with VC investment in the segment decreasing faster than VC investment in technology overall. While this data is admittedly a little coarse, and mixes telecom and data networking together, it nonetheless makes clear the direction such investments are taking. Indeed, as anyone investing in networking and communications will tell you, there just aren’t that many new ideas walking in the door. Many investors who made their living at layers 1-4 have moved onto cleantech, digital media and the online advertising economy. So can networking be made cool again?

It can, and here’s why:

M&A Landscape — Cisco catalyzed a significant change in the M&A landscape with the introduction of its Unified Communications System. By declaring open war on HP, IBM, and Dell, Cisco has made networking relevant again for the major OEMs who need a networking story — and subsequently turned them into potential startup acquirers. The market capitalizations of Juniper, F5 and Riverbed mean they could pay a price that would make the VC math work, and virtualization’s impact on networking puts VMware and Citrix into the mix as well. Three years ago, it was more or less Cisco or bust. Today there’s a long list of companies that could be home for VC-backed networking startups.

One part data center + one part cloud + one part virtualization=hot startup — Yesterday’s networking gear isn’t well-suited to today’s application workloads, traffic patterns and multitenant business models. Legacy architectures evolved from the LAN and designed for email and enterprise client server applications won’t cut it when it comes to those based on Map Reduce, nor will they work for collaborative filtering, putting together a Facebook page or renting CPU cycles by the hour. The data center has/will emerge as a distinct category of networking products. Legacy constructs like VLANs and subnets are wrapping cloud providers up in webs of complexity and management headaches, causing them to leave money on the table as servers fail to be fully utilized. Customers want SLAs but deep visibility into network behavior and performance remains elusive. Solving these issues will require not just new routers, but new ways of thinking about building networks.

Commoditizing forwarding — Much has already been written by James Hamilton of Amazon and others about the commoditization of networking and the router vs. mainframe comparison. Imminently, dense, non-oversubscribed 10GE switches will be available from ODMs for any OEM that wants to put a badge on them. Silicon and reference designs from Broadcom/Dune and Fulcrum will take new levels of price performance into the heart of the Cisco chassis switching lines. While they won’t have the full Cisco feature set, they won’t have the outrageous gross margins either. As this ecosystem emerges and technologies like OpenFlow and other efforts to bring the routing out of routers take hold, these advanced features may be provided by a community of third-party applications, decoupled from propriety chassis and operating systems.

Capital efficiency –- Building a networking company no longer requires $50-$100 million of venture funding. The appliance model is well understood in networking. An Intel Nehalem server can forward a lot of packets (PDF)! Perhaps more interesting, the virtual appliance model has emerged, meaning firewalls, load balancers and WAN optimization products can now be downloaded and run on hosts.

It’s ironic that as Sun Microsystems gets absorbed by Oracle, its long-held mantra — “the network is the computer” — has never been more true. For proof, look no further than social networking, search, SAAS and the cloud. And with them comes a host of interesting networking problems to solve. In other words, there hasn’t been a better time to start a networking company in recent memory. So what are you waiting for? Find some buddies and a whiteboard, and dream up something new.

Alex Benik is a principal at Battery Ventures

For more on cloud computing, virtualization and other related topics, join The GigaOM Network at its annual Structure conference June 23-24 in San Francisco.

10 Responses to “Can Networking Be Made Cool Again?”

  1. Dear Alex,
    While I do agree with other comments about the uselessness of new technologies to save a few percent, I’m more optimistic about opportunities for ‘reasonnable disruptions’ happening in networking. Re-thinking networks from the top-down applications/flows point of view (vs traditional bottom-up transport layer) change the rules in a quite fundamental manner. New techniques and concepts like Autonomic Networking and WAN Governance, as well as new deployment models, could well lead to the next wave (“beyond MPLS”). Cheers
    Thierry Grenot, CTO @ Ipanema Technologies

    • Thierry,
      Thanks for the note. I too am focused on what you call, “reasonable disruptions.” WAN performance, visibility, and cost reduction is very important with today’s distributed enterprise and the move to Cloud, SAAS, etc. Optimizing storage performance over the WAN is also an interesting area if long distance Vmotion is going to be a practical reality.

  2. Alex, I am not sure where you are going with this. The likes of Huawei and ZTE are already driving hard in the low margin market and I don’t see what value a startup can provide by discounting 20-30% off Cisco, Juniper, Riverbed and providing fewer features and inferior support to boot while still fending off the Chinese competition. I thought venture startups were supposed to be all about technology disruption and not about shaving a few margin points in price ? There just doesn’t seem to be enough room in that space in between Cisco and Huawei.

  3. “Commoditizing forwarding”

    Networking was cool because everyone could see what it was going to do the cost of communication, but somehow thought that they would be able to keep those sweet margins.

    The result communication (forwarding) became a commodity and VCs don’t invest in commodities. The returns aren’t worth it.

    The same thing will happen to cloud computing. Here is a way to do something cheaper and everyone sees it.

    But what is ‘cool’ here? The ability to attract VC funding?

    People may have forgotten the network in all the excitement of what you can do with it. But none of that would be possible with out the network. It has changed our lives for the better and I don’t think that ever stopped being pretty cool.

    • Allen,
      Thanks for the comment. I agree, networking has always been cool. Don’t read too much into the title, it was supposed to be tongue-in-cheek.

      Regarding margins and commoditization, networking isn’t commoditized, except perhaps at the low end Frys market. Networking vendors still enjoy sweet margins. Just look at the gross margins of Cisco, Juniper, F5, Riverbed, etc. Commoditization of forwarding is a good thing for start-ups and entrepreneurs. It allows them innovate in software above this hardware layer and not spend lots of time and money building chassis and forwarding hardware.

      Cloud and Web-scale environments are interesting because that’s where some of the toughest networking problems are. The scale-out server model has pushed complexity into the network. Virtualization then multiplies it. Some of these issues are about forwarding. Just getting enough bandwidth in place to cost effectively build non-oversubscribed networks at large scale is a challenge, but the more insidious problems are in software, in the control plane.

  4. whoop dedo

    why does networking have to be cool? it seems quite obvious that despite hype during the tech bubble (mostly a result of investor ignorance of what network players actually produce and market)…it seems obvious that this market would be a few large players making a lot of money selling metal boxes with blinking lights.