Cloud Hosting and Academic Research
“The system is fully implemented, but we can’t afford to make it public yet because of the hosting costs,” said Brent. I was asking Brent Hecht about his work on Ominpedia (my personal favorite project at the CHI conference this year). Brent is working on using caching to lower the hosting costs and hopes to make the system public, but his situation got me thinking. What are the advantages and disadvantages of hosting research projects on the cloud?
The advantages are clear: Eric Gilbert tells the story of his Link Different project unexpectedly going viral. His university server couldn’t handle the load. A cloud hosting setup would have handled this seamlessly. But then again, could his university budget have handled the bill?
My MediaMOO system was launched in 1993, and MOOSE Crossing in 1995. Believe it or not, both are still running. At one point I was going to close MediaMOO, but member Michael Day said he didn’t want to see it go away, and took over the server. When I moved from MIT to Georgia Tech in 1997, I bought a UNIX box to run MOOSE. Since then I believe it had to be moved to a new machine once–to a machine someone else no longer wanted that my IT department let me have for free. We bought the box, and we can just leave the server running on it. I reboot it once or twice a year (current MOOSE uptime: 5 months+). Just last week an ed tech researcher asked me some questions about MOOSE and I set him up with an account. He had read a couple papers and now he can actually try it out.
There is a value in keeping significant old systems around, even if they no longer have active user bases. A cloud hosting model seems so right to me–it’s scalable and robust. It just makes sense. But the hosting costs are a problem. Even if the total amount of money is small, grants are for specific work and have end dates. I can still be running a 10+ year old UNIX box, but I can’t still be paying hosting fees for a research project whose funding ended years ago, no matter how small that bill is. Grants end–there’s no provision for “long term hosting.” Our library can help us archive data, but they are not yet ready to “archive” an interactive system. I hope companies that provide hosting services will consider donating long-term hosting for research.
Developing research online systems, we typically work on short-term research grants. Those grants are time limited. It’s easy for me to put money for a server in a grant, and leave it running. The hidden cost is the burden on my university IT department, but little servers don’t cause a lot of trouble.
The Next Bison: Social Computing and Culture 
So, a few thoughts.
We have a study about to go live (in June). It is running on a department server, but it is too slow. So, we thought about using a cloud hosting service, but, then we got worried about IRB permission and security and the data we were collecting, which isn’t exactly medical data but is close (weight, activity, pain, sleep, etc). So, I just ended up buying a new server.
However, we were hosting a project that we did with Rutgers and U Wash years ago, and it was still being used some. It kept getting older and older, and eventually the version of Linux we had on it was no longer supported, so we couldn’t install patches. We could have moved it to a newer linux, but there were old versions of PHP, of Apache, of MySQL, and some more specific tools, that we would have had to do a lot of work to get it moved. And the grant was long expired and spent, so we just turned it off one day.