Leon Woo's Explanation | 3Com: Creators, Choi

Leon Woo's Take

Leon Woo’s Explanation of the Technology Race

Leon Woo, founder and VP Engineering of Synernetics, provided some insight into the competing technologies at the time 3Com acquired Synernetics and afterwards. He noted in an email to author Jeff Chase:

We had a 3Com-killer in development that was based on ASIC technology that we revealed to 3Com in due diligence that caused 3Com to double the acquisition price. At that time, BICC Data Networks had started an ASIC development as well to switch Ethernet. 3Com, to better position the products, had BICC division package their silicon into “stackable" and Synernetics division package their technology into "chassis". Regardless, ASIC based switch was key. There is a story of the BICC ASIC being from DEC’s manufacturing team in the UK. They were highly regarded...

I think I wrote a tech sequence for you in my first review on the timing of the industry’s development of fast Ethernet, which killed FDDI. It created one of my life lessons, “Never bet against Ethernet,” as we embarked on Gigabit Ethernet switching silicon along with the BICC group… The key acquisition in the Ethernet switching race was really Cisco’s acquisition of Grand Junction (not Crescendo or Kalpana as most people think). Grand Junction “invented” fast Ethernet and was ironically founded by Ex-3Com people. In my opinion, that tech acquisition allowed Cisco the reach escape velocity much faster than 3Com, Wellfleet, Synoptics and Cabletron (the big 5 networking companies in the 90’s)...

What happened was that in the original product (the LANplex 5000 we OEMed to 3Com, renamed Linkbuilder 3GH), the backplane in the LANplex Product was FDDI (which was the standard before Fast Ethernet was standardized). Both Microsoft and Boeing (our first large campus deployments) used FDDI as the backbone. In most cases, switched Ethernet to the desktop. Bear in mind that Fast Ethernet was not yet invented/standardized (with competing standards for Fast Ethernet by Grand Junction/Cisco and HP). It was a technical challenge to switch Ethernet onto an FDDI network because it required a packet header translation and packet size fragmentation/re-assembly (FDDI had a bigger packet size than Ethernet).

The standard “backbones” in the LAN in the early 90’s was FDDI and the upcoming ATM networks (recall ATM was the reason 3Com bought Nicecom). Fast Ethernet came onto the scene in the mid 90’s and demolished FDDI and ATM in the enterprise. ATM was relegated into the telecom marketplace (Cascade Communications was the big winner here) competing with Frame Relay. FDDI had a foray into the metro networks, but essentially completely faded away.

Backplane technology and Backbone technology are different notions. High speed backplanes (almost always proprietary) interconnected ethernet switches inside a chassis to scale the port count for large enterprises. Backbones are standard high speed networks (like FDDI/ATM/Fast ethernet) that connected network boxes and servers together. Realize now that Gigabit Ethernet, 10GE, 40GE and now 100GE supplanted FE as a backbone technology and 10/100/1000 Ethernet became the desktop connection. Synernetics had a high speed “backPLANE” under development that 3Com would not have access to in addition to wire speed switching ASICS (which was the main reason 3Com acquired Synernetics). Synernetics was always fearful that 3Com would design us out, so we heavily invested in ASICs and “BackPLANE” technology. It turns out we were correct given the ASIC developments at 3Com’s BICC division in the UK.

To scale Ethernet large port counts for large data centers and enterprises, there were two products in the industry – Chassis’s (with “proprietary” high speed backplanes connecting switch cards) and stackables (with a “proprietary” high speed stacking connection between stacks). The choice by customers in deploying Chassis vs. Stacks was largely religious in the early days.

Like Cisco, 3Com tried to hedge technology bets by buying companies with ASIC developments in different network technologies. Like NiceCom for ATM and Chipcom (who actually had some FDDI developments). I did the due diligence on those two ASIC developments and neither ended up working reliably in the marketplace. It was always more difficult to switch Ethernet to ATM, FDDI, or any other network technology for the reasons I mentioned before (packet header translation, packet size fragmentation/re-assembly). The packet translation was not needed in switching from Ethernet to FE to GE to 10GE to 40GE and so on. We rapidly realized this after FE came on the scene and basically took over the enterprise away from FDDI/ATM. The only standards that survived were the niche technologies developed for Storage SAN’s – fibre channel and the ultra high speed network called Infiniband (which is competing with 40GE today). So “Never bet against Ethernet” survives to this day with the latest challenger being WiFi to the desktop and perhaps now 5G in the carrier space!

Leon