Tell us a bit about yourself/team
As our website http://bigboards.io indicates, we are 4 professionals with various backgrounds in software and technology. In the past years we all have become firm believers in the possibilities and opportunities that Big Data can offer. However, we all look at it from a different angle: business, marketing, software stacks and hardware. But that actually makes us a stronger team.
What project are you currently working on?
Our experiences in the field of Big Data convinces us there is an enormous lack of a good educational platform to learn Big Data technologies. Note that these are intrinsically all cluster technologies, hence, you need multiple computers to truly experience the implications. You not only have to make an extra financial investment to buy the hardware (nodes, power, network, cases, cables, etc) but also (and moreover) the additional investment in time because of the extra complexity in the software setup, deployment and management. We experienced these issues hands on when we were running hackatons at bigdata.be for enthusiasts wanting to learn Big Data technologies.
For multiple years, we have been toying with various ideas in diverse areas. Finally we hope to bring a lot of them together in BigBoards:
• A big data cluster should consist of a number of identical nodes (obviously)
• One big data node should be only a sandwich of a motherboard and a harddisk to bring the processing as close as possible to the data
• Use ARM technology to reduce the power consumption of one node
• Scale out a cluster instead of scaling up a single node.
• The ratio of reduced power consumption per node and the number of nodes in the cluster will favor the power consumption of the whole cluster
• Use a chimney design to bring multiple nodes together. Use natural convection to cool down the nodes.
• After some first 3D designs, we naturally came to an hexagon layout with an integrated network switch and power distribution as bottom- and backplane. (Cfr. sketches)
How did the Wandboard help you to kick-start your project?
The idea behind BigBoards came rather quickly, but we still needed to validate the idea. We wanted to create a prototype, install a big data platform on it and start doing some tests. Since we do this in our spare time, we have a limited budget so we were very glad when we found out about the Wandboard. We quickly ordered 6 Wandboards and a few days later we got started on building the prototype. The result can be viewed at http://ow.ly/i/2R9tw
Why you picked the Wandboard and how does it fit in your project?
BigData is all about bringing processing close to storage. Because of that we need an energy efficient board which still has plenty of memory and most importantly, a SATA connector to connect a 2.5” harddisk. We have searched a long time for this combination but the options that we found either had insufficient memory or lacked a SATA port. Eventually the Wandboard fits most of the needs. To be honest, we could do with more memory, but nevertheless the Wandboard really gives you value for your buck.
The Wandboard is completely opensource, is this important for your project?
We were not specifically looking for open source hardware when we started out with BigBoards. Actually, we barely knew about the fact that hardware had been open sourced. But having access to the designs, especially the 3D ones, really helped us to create the 3D printed parts we needed. No need to measure everything, just take the 3D design and engineer the custom parts you need.
You can imagine that we will dive into the schematics of the Wandboard a lot more in the next steps of our project. Having the possibility to do so really, speeds up our work
What are your next steps and will you use the Wandboard in your final design?
So, after buying the 6 Wandboards, we 3D printed some supports (cfr sketches) to bring the Wandboards and harddisks together in our hexagon layout. We also bought some temperature sensors, which we hooked to some Arduino’s. That’s our continuous data logger.
Right now we are setting up our lab prototype to have it run a perpetual data sort which continuously pushes out it’s measurements to the internet. That way we will create a live stream on our site. The goal is to prove that this setup can actually run big data workloads and that the chimney design is efficient enough to cool the various components without the need for active cooling.
We would like to use the Wandboard EDM modules in the final design since they have a small footprint and allow an easy upgrade path when more powerfull modules come out.
How can people that have a Wandboard join your project?
We try to open up as much of the designs and tools as possible. This way we hope other people can benefit from the things we try out and experiment with. In a later stadium we will bring out software packages called droplets which we will have to test. Of course feedback, comments and ideas about BigBoards are always welcome.
If you have a new project. Will you use the Wandboard again?
We are really charmed by the Wandboard and it is certainly at the top of our head when we would do a new hardware related project.
Where can people get more information about your project and get in touch with you?
Very soon we will be selling our prototypes on demand that are built from standard components (incl. 6 Wandboards), a 3D printed skeleton and a custom case.
All information of our progress and final product is available at http://bigboards.io