A History of the Server Blade

UPDATE

Chapter in the history of Blade Computing closed

On October 3, 2005 HP announced an agreement to acquire RLX Technologies™, Inc., an early pioneer and leader of blade hardware and management technologies.

RLX Technologies, Inc. Company Profile


The Beginning

1999

The idea for server blades originated from observing the need for an easier and more economical way to deploy large numbers of rack mount web servers in data centers. In the late nineties, I was working with customers in the mass hosting and co-location business who were becoming increasingly frustrated with the hassles of deploying large numbers of 1U “pizza box” style servers. Many of them were installing large numbers of Cobalt RaQ (now Sun/Cobalt) servers, which were a great solution at the time, but still had their own issues. It was then that I realized that there was a market for a more efficient hardware/software platform and better tools for managing them.

It was amazing that data centers could operate with such wasteful power consumption, horrendous cooling and cabling, and lack of reliability. The deployment and management of these servers was becoming a headache of catastrophic proportion. What had happened was that while tier one vendors were busy one-upping each other by cramming hotter CPUs into smaller and smaller sheet metal boxes, they completely forgot about efficiency! It was obvious to me that making servers smaller, while simultaneously increasing CPU megahertz and thermal output, was not a sustainable trend. There had to be a better way.

At that point in time, there were no research reports predicting that power consumption, heat or complexity were going to become key issues in data centers. And, there were certainly no market studies showing that blades were the next major trend in computing. Ever more powerful pizza boxes, even some 4-ways, were all the rage. However, I could see that frustration with hot, complicated and expensive infrastructures presented a market ripe for a new solution. It had to be simple, low cost, reliable, as well as easy to deploy and manage. It was clear that when a market that was growing the way it was and the choices so limited, that the time is right to introduce a new product concept.

Finding nothing available to meet the this need, I began by taking a step back and looking at how we arrived at that point in 1999, and tried to visualize what would be the market drivers for the next wave of computing. What was immediately clear was that scale out architectures were coming, and would over time replace the expensive and inefficient monolithic hardware that had dominated the data center for the previous 10 years. This, along with the ever increasing adoption rate of Linux, clustering, and commodity x86 hardware, would all need to be considered when setting out to create a solution to data center inefficiency. It would need to be a completely new server form factor that would eliminate all unnecessary heat, sheet metal, cables and components. Basically, eliminate anything and everything that wasn’t absolutely necessary to the task of serving up web pages. The early sketches on napkins were not about blades, but more about consolidating, simplifying and most of all, lowering power consumption. Instinctively, I knew that finding ways of lowering power consumption without sacrificing performance, would be the key. It would be those early scribbles that would eventually lead to the invention of the blade architecture and would turn out to be the most elegant way to address the need for efficiency.

Little did we know at the time, but we would be creating a ‘disruptive technology’ before ever having heard the term, and certainly before reading either Clayton Christensen’s The Innovator's Dilemma or Geoffrey Moore’s Crossing the Chasm.

Throughout 1999, I had been consulting part time for my friends at Cracken Harkey, Co., a Dallas based investment firm. I was assisting in evaluating various dotcom start-up business plans for possible investment. All the while, I was continually nagging them about the need for an efficient web server. In late December, after a lot of selling on my part, they were ultimately convinced that internet infrastructure, and in particular, web servers, could be a solid investment strategy, and the new company, RocketLogix, was born. We began pursuing on the efficient server concept on January 1, 2000.

2000

The original designs I had been considering all centered around mobile Celeron processors from Intel because of their low power consumption and relatively good performance on web benchmarks, but when Transmeta announced the Crusoe microprocessor on January 19, 2000, I realized that it would enable us to double the density of our design. The efficiency of Transmeta's TM5600 processor easily supported an ultra-dense design of 24 blades in a 3U chassis ­ 8 times the density and one sixth the power consumption of the standard 1U web server of the day ­ at the same performance level on web serving applications! This was truly a major breakthrough.

In March, we went into the market to validate the business proposition and found that the ultra-efficient, ultra-dense, blade computing message was right on the mark. After an exhaustive period of market research and feasibility studies, we began electronic and mechanical design in earnest. I worked hand in hand with the our mechanical and electronic designers, solving many complex problems that arose during the design and layout of all the various boards and sheet metal that make up the system. Although it was an elegantly simple design, nothing like it had never been done before and therefore it had design issues that had never before been encountered.

In April we raised a million dollars from friends & family and kicked off hardware prototyping. And in June 2000, we moved into a new headquarters in Plano, Texas and hired the first few employees. Up to this point, the company had consisted of a core group of consultants that numbered only four or five, myself included. In July, we received our first working prototype from our outsourced design house. At this point the company consisted of 15 people and we had filed 6 U.S. Patents with a total of 187 claims related to dense blade computing. With working hardware in hand, we were able to demonstrate the blade concept to potential investors, and our second seed round closed in August at $5.5M.

By Q4 2000, we had raised an additional $13M, and the blade revolution was officially underway. OEM discussions during September and October revealed our technologies to potential competitors and we attracted an executive team by end of year.

2001

In January 2001 the new team was on board, the company became "RLX Technologies, Inc.", moved to the Houston suburb of The Woodlands, and a Series A round of funding would be closed in February for $40M.

That spring, the phrase "server blade" entered the lexicon of computer geeks the world over and a few small competitors appeared on the horizon. From the beginning it was clear that the big companies would not react quickly and thus we had a unique window of opportunity to gain the proverbial beachhead.

To that end, we drove to release a complete, bug-free and vapor-free product, which we did in May. RLX Control Tower management software was released shortly thereafter. The products were received with acclaim and the blade concept widely recognized as revolutionary. RLX was named one of Red Herring's Top 10 Companies to Watch in 2001.

With the market changes of 2001, the beachhead proved elusive but RLX drove product improvements and releases on an accelerated schedule. By year's end, RLX began to shift direction in response to demands for higher performance in markets outside its original target market. Despite this, throughout 2002 RLX continued to stay ahead of the competition due to the strength of its core technologies and maturity of the product.

2002

In 2002, the industry began to spawn blade-related forums and organizations. In April, the first trade show, Server Blade Summit attracted early adopters and those interested in guiding the growth of blade technologies as an industry. The new Server Blade Trade Association was formed at the close of the year and I was invited to join as a founding member.

In May, the Los Alamos National Laboratory project called "Green Destiny" was unveiled. Green Destiny is a compute cluster made up of 240 RLX blades that demonstrates that the supercomputing world is also interested in efficiency, reliability and availability. The unveiling was attended by Gordon Bell and Linus Torvalds, both of whom saw the potential of ultra-efficient cluster computing when we were pitching the concept back in the year 2000.

Finally, in late 2002, tier one competitors began entering the market. HP, Compaq, Dell and IBM were all shipping blade products by the end of the year.

In 2002, I realized that the potential of the efficiency message had yet to be fully realized and would be best pursued apart from RLX. Trends continue to point to resource conservation as the key issue of the future. Market instability and global uncertainty reinforces the need for efficiency as does the existence of organizations such as the Rocky Mountain Institute that work to drive adoption of green technologies.

2003

The year 2003 started with the Sun announcement of its first blade product. A design that remarkably resembles the original product we created at RocketLogix some three years earlier. The company that acquired Cobalt Networks for $2 Billion in 2000 has abandoned the 1U market in favor of blade servers. In just two short years, blades are becoming a ubiquitous component of server technologies and next generation computing solutions.

There are many markets that today face the same issues faced by my customers in the late nineties. These markets can now take advantage of new, revolutionary technologies or new combinations of existing technologies. In some cases they might be blades, but then maybe not.