FS3 Project: Alpha Boxes and Ad Hoc Project Issues
Intriguing developments occurred in the past few days on both the equipment front and the ad hoc tech project aspects of the FS3 streaming supercomputer project.
In the past eight days I've been in four meetings for the FS3 project. A weekly teleconference was established to monitor project progress, make necessary adjustments and discuss new issues. Two of the four meetings were the weekly teleconferences.
The third gathering was an organizational project meeting on 18 February in Green Bay, Wisconsin, USA. The FS3 was explained to people interested in it who had not yet had a chance to extensively discuss the project. Numerous aspects of the project were covered ranging from "where the heck did this project come from" to "should the FS3 use gigabit interconnects, 10GigE, or something else?" The meeting lasted two to three hours, but would have gone longer if people hadn't had other commitments. Three meeting participants reconvened at Lin's Garden after the meeting for some tasty Chinese food and to discuss the supercomputer and a plethora of other topics.
During the organizational meeting on 18 February, the consensus was reached to add an alpha stage to the FS3 project. The alpha stage goal is to build a small cluster of computers for demonstrating the streaming supercomputer concept at a relatively low cost. FS3 alpha is targeted to have 6 boxes with 12 GPUs. Justin put together specs for the important components on Sunday night, 18 Feb. Today I put together detailed parts list spreadsheets for two approaches to the FS3 alpha. One is a middle of the road cost and the other is a bare-bones low-cost system. Both approaches have the same GPUs, motherboard and RAM, but have differences like two 500 GB HDs vs one 250 GB HD, a UPS vs a surge suppressor, etc.
The FS3 alpha concept is that a 12-GPU system gives an entry point for people to play with the hardware and see what they can make it do. After building the boxes and installing minimal software, the project team will take the FS3 alpha on road trips to universities and tech events. This gives the project some exposure, creates buzz and gets new people intrigued by the concept to the point where they'll work on the project or ask how soon they can buy an FS3 fifty-GPU system!
Now that the alpha system is spec'd out, the last two steps before buying the hardware and starting the snowball rolling downhill are securing funding and establishing the FS3 company (governance structure, equity or ownership structure, LLC vs S Corp vs ?). The team is working on those issues as this post is being written...
The fourth recent meeting on the FS3 project was a 6 a.m. IHOP breakfast on 19 February. I met with one of the project advisors, Kurt, who posed important questions and offered great advice. Had he not had an early morning teleconference scheduled, we might have sat there, drank more coffee, talked and worked on the project for another hour or two. One of the Kurt's suggestions is playing out today and over the next few days. That suggestion is to get the basic ground rules laid down for the project. The four ground rule areas we discussed on 19 Feb are:
- Define and put in writing the governance and ownership rules for the LLC, or whatever organizational form is chosen.
- Define how the company will recognize and make the transition from a 'research project' to a capitalist company with the goal of making a profit.
- Define how FS3 intellectual property will be approached, protected and owned.
- Develop expectations for all project members, clearly spell them out in writing, get alignment on those expectations and manage expectations throughout the project.
- Help build the culture of innovation in the state.
- Enable or enhance advanced projects at universities and other research organizations in the state which require low-cost HPC resources (high performance computing).
- Enable advances or breakthroughs in alternative energy and biomedical/biotech, two recognized focus sectors for Wisconsin's future.
- Enable some awesomely cool games, hopefully leveraging collaboration with the four Wisconsin video game companies.
- Issue meeting or discussion notes on the day of the meeting or discussion. Take the time -- MAKE the time.
- All meeting/discussion notes must include action items or next steps agreed to in the meeting, listing the item, the person responsible for completion and the target date. Those agreed-to expectations need to be followed up on and managed by the project manager.
- Any ad hoc tech project needs a simple, compelling, professional website to which non-project people can be referred. The FS3 project does not yet have such a website, and its lack was felt when a BFG Technologies employee asked me this morning if we had a website. Arghhh! We need to get the content written, then build and launch the website. Soon.
- Ad hoc tech project members will have hugely varying agendas, expectations and levels of expertise at communicating and collaborating. The important thing for people to focus on in these projects is continual improvement. Mistakes WILL be made -- progress will be mighty slow if none are expected or accepted. But the focus will be on what we learn, the advances we make and the passion we have for what we're doing. Those who wish to focus on mistakes will do better in a large corporation than on an ad hoc tech project team.