I worked in the MicroTCA industry when it began in 2006, and I stayed involved through 2009. Now, as an outsider, it is interesting to compare what we anticipated at the industry’s inception to what has been revealed by the market.
In 2006, MicroTCA arrived as a low-cost system architecture for a variety of industries such as telecom, enterprise, medical, and digital imaging, among others. Industries that had a high computing requirement for the lowest possible cost were ideal candidates for MicroTCA. The system’s modularity allows components to be shared among industries, which in turn drives up the volume. High volume – in theory – would drive down the cost. The PICMG organization led the standardization of MicroTCA. Two of my previous employers – Artesyn Technologies (later acquired by Emerson Network Power) and MicroBlade – were very active with PICMG in defining the standards. In addition to its standards work, PICMG set the guidance cost at $500 for a small system at high volume (100Ks systems and up annually) within five years of its introduction. It was a very aggressive guidance cost given that the system was yet unproven.
The subsequent years, 2006 through 2008, saw the MicroTCA ecosystem grow with established companies such as Motorola, Artesyn Technologies/Emerson Network Power, Radisys, Kontron, along with smaller companies and startups – CorEdge, MicroBlade, NAT, Performance Technologies, SanBlaze, and many others. Fast forward to 2011, and several startups were gone (including CorEdge and MicroBlade). Embedded system powerhouses de-emphasized MicroTCA development. For example, Emerson Network Power acquired Motorola Embedded Communications Computing Group – a MicroTCA and embedded system group – in 2008, and no new MicroTCA-related products have been introduced since. A search on the PICMG website in April of 2012 shows 69 product listings with Samway Electronic that seem to be a pure play of MicroTCA. I personally know that several of the products in this listing are no longer in production. A quick search on Google News reveals that MicroTCA-related news hit its peak in 2006 with three pages of results and dropped off since with an average of one and a half pages of results. Google News did not return any MicroTCA news in 2011.
Flexible, modular system architecture normally carries some price penalty because they inherently need more parts. Trade-offs are made between features and cost, which is always a balancing act. In my experience, Original Equipment Manufacturers (OEMs) – our customers – demand flexibility with a minimal cost penalty. MicroTCA’s arrival felt like a savior to this ever-increasing demand, especially with the promised $500 system cost. In practicality, MicroTCA was overdesigned – high speed backplane, challenging thermal management, and complex modular cards. Early adopters were shocked by the sticker price. In a similar example, I recall trying to build a low-end VoIP server in 2008. I needed a processor card, a storage card, along with a switch and a chassis. I could only locate an Intel Pentium-M processor card. 2008 was when Intel discontinued the Pentium M processor line. The system came to about $20,000. Meanwhile, a low-end server from Dell was about $1500 - $2000 with a much more powerful processor. It was an order of magnitude more expensive to build it with modular units. Granted, MicroTCA was still in its infancy and needed volume to drive the cost down. However, the initial cost would need to be somewhat reasonable. Imagine if the first Toyota Prius were priced at $200,000 – the hybrid vehicle market would never have gotten off the ground. Moreover, MicroTCA would have to drop its price three times faster than competing technologies (assuming that the competing technologies would reduce their price on an average of 10 percent annually) with any hope to catch up 10 years later – again assuming that MicroTCA reached a certain volume each year to warrant an aggressive price drop.
I like MicroTCA for its modularity and flexibility, though it seems the complexity and high cost will prevent a wide adoption. A few of us proposed a simpler version of MicroTCA shortly after its introduction, but to no avail. Until MicroTCA is significantly cheaper and simpler, it will only be a great system architecture for niche markets.
About the author: Thanh Nguyen spent over two decades in embedding computing as an engineer and product manager for a variety of companies including IBM, Rockwell, Emerson, and MicroBlade. He also attempted to start an open source hardware company for MicroTCA. He is now at Johnson Controls focused on battery system R&D. He has a BS in Electrical Engineering from Penn State University and an MBA from Regis University.
Edited by
Rich Steeves