Stealth Marketing, and the CEO
Imagine a corporation leading the High-Performance Computing (HPC) networking market in 2006, producing $30 million in annual revenue and doing it with only 49 people. Meanwhile, the CEO is spending valuable time hand-editing, in HTML 1.0, the company’s website. That was Myricom. The CEO, at that time, was the brightest mind in networking that I’ve ever met, but he also had several serious blind spots; for example, he firmly believed that the sole role of marketing was to issue press releases, and there was a contractor for that.
In 2006 Myricom began shipping its first production single port 10G Ethernet NIC based on their network controller ASIC, which came out of the oven in 2005. This chip was architected to manage both Ethernet and Myrinet 10G packets, so it was a transitional controller, and with it, we produced an affordable single-chip NIC. These were the early days of PCIe, and the first generation was fragile, so adoption took time. Server motherboard vendors back then had just begun offering PCIe, and a single eight-lane connector was often all there was on a motherboard. By definition, a first-generation eight-lane PCIe slot delivers 16Gbps in each direction before overhead. Therefore, under the right conditions, a well-designed single-port 10GbE card could achieve line-rate data to host memory. Myricom had produced just such a card, but outside High-Performance Computing (HPC), nobody knew it.
In 2007, Myricom turned the crank and produced the second generation of their network controller ASIC, which was engineered to offer two active Ethernet ports. Unfortunately, someone made the horrible decision to pass on using a PCIe generation two IP block for the chip, so it remained as a PCIe generation one ASIC. Unfortunately, the PCIe eight-lane bus could not sustain wire-rate on both ports. Myricom could have produced a very effective single-chip, dual-port NIC with this ASIC. Granted, it couldn’t support wire-rate on both ports simultaneously, but for performance-sensitive applications, it could have at least have played a role by providing an active hot fail-over. Unfortunately, our HPC heritage and corporate culture wouldn’t permit creating a dual-port product where the second port couldn’t support wire-rate.
Later that year, Myricom put two of these new controller chips, along with a second-generation PCIe switch chip from another vendor, on a card, and we delivered their first dual-port wire-rate 10GbE NIC. That silicon and board configuration is still for sale today, but back in 2008, we were still having trouble gaining traction in the market. The primary reason we had difficulty selling 10GbE products was that there was ZERO marketing. So on my own, I spun up the 10GBE.net site to act as a buyers guide to 10 Gigabit Ethernet NICs and switches. Throughout 2009 the 10GbE.net site saw thousands of monthly page views, and at one point, it was referenced by The Register as the authority on 10GbE.
Other than my wife, only two people knew 10GbE.net was my site. One was Myricom’s midwestern sales director, and the other was a friend who was a director of HPC sales at IBM and the person who recommended me for my sales position at Myricom. I swore both to secrecy as they knew the ramifications for me if the word had ever leaked out. As the site admin, I would regularly receive email questions from corporate executives at both potential customers and competitors. Over the summer, my CEO wrote to the site’s admin, asking various questions and thanking me for promoting his company’s products. I responded to all his questions but never revealed my true identity.
In the late fall of 2009, Myricom attended its primary trade show of the year, Super Computing 2009 (SC09). Myricom had been attending SC since the beginning, so their 20'x20' booth always had a premium location. During the show, my friend from IBM was strolling by our booth, and the CEO called him over. We had a pair of 60" screens in the center of the booth scrolling slides. The CEO brought up a browser and then loaded my 10GBE.net website. He then spent 30 minutes pitching this IBM executive on our products using my website. Ten minutes into his pitch, the CEO calls me over to drive a few other points home about the site. It was all we could do for my friend and I to contain ourselves during the CEO’s monologue. He had no idea that we both knew it was my site.
While the 10GbE.net site was an awesome tool for marketing and positioning Myricom’s boards, running it in stealth mode for well over a year was taking a toll on me. Soon after SC09, I took the site down and parked the URL. By fall 2010, the then COO overthrew the CEO and founder to become the new CEO, but how that unfolded is another story for another day.