How to Design a Data Center for the Developing World
Note: Inveneo, in partnership with ARM Limited, LeMaker, and Protocase, is pleased to announce that a winner has been selected for its recent $10,000 Micro-Data Center Design Challenge. Congratulations to William Weatherholtz and his team “Micro Weather” who are the winners of the $10,000 challenge! Coming in second place is “RuggedPOD”, a team from France who was lead by Jean-Marie Verdun. Each member of RuggedPOD will receive a Google Nexus 7 tablet as a prize for his or her hard work and design.
They have different interests. One teaches dance. Another is a helicopter pilot-in-training. One is a software engineer. Another is a carpenter. Still another, a studio artist. But they came together with a shared purpose: to try to improve conditions in Third World countries.
Meet William Weatherholtz and team, who just won the Inveneo solar-powered Micro-Data Center Design Challenge ( Bruce Baikie) for their Micro Weather station design. The team’s winning entry is an object lesson in how creative methodology, a diverse team and carefully considered components selection just might help transform developing societies.
“I have a soft spot for Third World countries and I’m really interested in finding ways to improve conditions there,” Weatherholtz said in an interview. “I felt like this was a project that played to my strengths and my desire to educate.”
The design criteria for weather stations is unique: How do you deal with rain, rust, long-term durability, a lack of power sources, and little critters that like to gnaw on things in the wild? Weatherholtz (pictured to the far right of the nearby photo) and his team (pictured, L-R: Garrett Johnson, Victoria Johnson, Kelly Weatherholtz; not pictured Joshua Wickern, Bradley Weatherholtz, Landon Weatherholtz) embraced a unique methodology that included using Edward de Bono’s Six Thinking Hats philosophy. The approach is designed to help improve team perspective and collaboration during projects. This was a particular interest because the seven team members were dispersed across the country.
“Everyone was assigned a different perspective,” said Weatherholtz, a mechatronics engineer. “So for example, someone was assigned an aspect of the design that only considered price; someone else would focus on aesthetics, and so forth.”
The team rotated through these different considerations and perspectives and then amalgamated different parts of the design into the one they liked.
The team started by identifying the customer needs and translating those into engineering characteristics:
- What type of battery was required?
- How much back-up power would be needed? (the team targeted five days for it to run on back-up power initially but ended up at 2.5-3 days—more on that shortly).
- What other design considerations might be unique for a developing country?
- What were the environmental needs of the device casing?
Here’s a look at how the team tackled some of the design considerations.
This was an extremely critical component that needed to be as reliable as possible. Additionally, the team had to understand how much power they could pack into a small space. Should they push the limits for longer back-up power capability and accept the consequences? Additionally, what type of batteries could be shipped internationally?
“We tried to pick a battery with a very high energy density and moderate size, but the battery is still pretty heavy and large,” Weatherholtz said. “Adding another battery would mean another cubic foot of space and an extra 60 pounds in the design.”
At the end of the day, two and a half days backup capability seemed good enough for most applications, he said. That meant the battery could recharge in four hours with sufficient power, and most places get at least five hours of good sunlight, he added. The team ended up selecting an absorbent glass mat (AGM) battery—essentially a golf cart battery—that doesn’t spill, tip or have vulnerable components inside.
This was one of those developing-country considerations, where ready reliable power sources are hard to find, if not non-existent. Even though it was the team’s first time working with solar, adopting the technology was key. “It’s a fantastic solution in a remote data center application because a data center is a static structure,” Weatherholtz said. “It allowed us to take advantage of that big fusion generator in the sky.”
The team considered plastic but wanted the system to be able to take a pounding. So they settled on aluminum, a reliable material, which conveniently could serve as a sizable heat sink. They designed to a worst-case scenario of 50C ambient temperature with direct sunlight, no humidity, and no moving air.
“One of our main design criterion was to make the enclosure—and enclosed electronics—reliable. For us, that meant it needed to be completely sealed with no moving parts,” Weatherholtz said.
Single Board Computer
The contest criteria specified the SBC. As a designer and engineer, Weatherholtz said he doesn’t really like being shoehorned into a solution, but, that said, “the Banana Pi boards were hard to beat,” he acknowledged. The Banana Pi, based on ARM Cortex-A7 with Mali-400 GPU and running open source software, is designed to be inexpensive, small and flexible.
The technology was “robust, open source and low power,” he said. “When you’re dealing with IoT applications and micro data centers, you don’t have a lot of power and you can’t have a lot of heat, so ARM is best.”
One challenge is that boards such as this typically have two sources of heat — RAM and processor. The team undertook considerable thermodynamic analysis and determined that getting rid of heat was key. The Banana Pi boards were ideal, Weatherholtz said, because the two sources of heat were on the bottom face and as a result, the team was able to direct the heat in the optimal direction. Had the CPU/RAM been on top, then it would have been more challenging to get the heat out, he added.
Weatherholtz and team spent a total of 150 engineering hours on the project, for which competitors used ARM-based solutions to create the “micro-board chassis” designs. They will share the $10,000 prize and the design will be built and deployed in the developing world.
“I really can’t overstate the importance of thermal analysis in projects like these,” Weatherholtz said. “If heat doesn’t have a good way to escape, it’s going to build up and cause high temperatures that make your electronics fail, or at least fail prematurely.”
“For us, making a low thermal-resistance path out of the case was a main design consideration. We identified where the heat was being generated (see image right), and then got it out. Everything centered on that. Where we placed components, what we placed them on, how we connected them to what they were placed on… everything.”
Written by Brian Fuller, an employee of ARM. This post was originally posted on ARM Community and was republished with permission from the author.