Author: Jay Lyman
To the projects it hosts, OSL is an integral part of daily development and downloads.
“Gentoo wouldn’t be where it is today without the support of the OSL,” says Gentoo Linux Board Member and Infrastructure Lead Kurt Lieber. “They’ve been a long-time supporter of ours, offering free hosting, bandwidth, use of hardware, etc. They’ve also established a very robust, scalable mirror infrastructure with ~1Gbps of capacity. We rarely have download issues now when we release new versions of Gentoo.” Lieber says OSL services have been exemplary, “and in fact, it’s better than what I would expect from commercial vendors in a lot of respects.
“They approach things from the standpoint of, ‘how can we help Gentoo,’ instead of ‘what are we obligated to do under the terms of the contract we negotiated,'” Lieber says.
Drupal founder Dries Buytaert says his project’s hardware — two dedicated Web servers, one dedicated database server, and another email/CVS server — was provided by Drupal and set up with the help of the OSL, which provides bandwidth and power. “OSL helped us architect the setup, and installed all the machines,” Buytaert says. “In addition, and this is invaluable on its own, they help us with the day-to-day maintenance of the Drupal infrastructure (e.g. backups, upgrades, troubleshooting). You know, it is still hard to believe that they give us such a remarkable service … for free. It isn’t very often that one gets offered a deal like this. It is safe to say that OSL is one of the Drupal project’s key contributors.”
So how did this open source hosting effort come about, and what kind of infrastructure does it take to handle the bulk Linux kernel discussion, Firefox, and other downloads? OSL Associate Director Scott Kveton, with help from his team, explains, beginning with the lab’s origins:
Kveton: We had a spend-crazy CIO who ended up leaving us for the East Coast about eight years ago. OSU’s IT department was left holding the bag. We were $6 million in debt with a $4 million annual budget. The answer was for the University administration to curtail our spending by $1 million a year until we paid down the debt. It (obviously) took six years.
During that time, our enrollment was going from 13,000 to 19,000 and the use of the Internet and network on campus was skyrocketing. We had increased demands and significantly more resources to deal with them. In 2001 I returned to the University from stints in the dot-com world. I was just a system engineer with a hobbled IT budget and lots of things to do. Since we didn’t have money for all of the Sun gear we traditionally used, it was easy to make the case to go to Linux on commodity hardware (Debian on Dells).
Long story short (and this is me not being brief), we looked to open source for all of our core infrastructure needs because we didn’t have a choice. I was able to replace my entire infrastructure with commodity gear running Linux for a quarter of my annual Sun hardware budget. Five years later and all of that gear is still in production (not maintained by me anymore). During all of this, OSU maintained an open source mirror at ftp.orst.edu. In 2001, when I returned, we had 72G of RAID5 disk space connected to a SPARCstation 20. Yep, a Sparc20. We mirrored a few Linux distributions, CPAN, and the id software archive.
NF: How did you learn about the need among open source communities and projects like Mozilla, Debian, and Gentoo for Web hosting?
Kveton: It really came out of the mirroring we were doing. We had a relationship with them through helping them upload their bits to our mirror, and we always prided ourselves on doing a good job with that, even with the limited resources that we had. In 2002, Zack Welch (former Gentoo developer and Corvallis resident) came to me and asked, with all of the use of open source we were doing, wouldn’t it be great if we could host some of these projects too? Zack kicked in two Shuttle XPC boxes that we hosted on a bread rack, which were used for Gentoo developers. I want to say CVS, shell, and maybe the Web site were on there. Gentoo led to Freenode, as some of the Freenode admins were Gentoo devs. Soon we helped Debian out, again through our mirroring interactions with the hosting of a few development servers. Then Mozilla was spun out of AOL and was pretty much without much of an infrastructure. We started with managing their mirror network (keeping people up-to-date, notifying them of next releases, stuff like that) and that led to a bigger relationship in terms of them landing their first boxes here. They sent a few Dell 2650s, which hosted some of their new Web services (what would become the application update service and addons.mozilla.org).
Freenode led to Arklinux, Debian to Software in the Public Interest. PHPBB and Drupal turned to us through people in the community involved in other projects. Kernel.org turned to us when part of their existing hosting infrastructure was going away; we were the obvious choice because nobody else was doing it and we had a good reputation with the community. It’s been extremely organic and we’ve always focused on being able to handle the needs of our customers on a daily basis.
It should also be noted that our team here at the OSL does much more than just Web hosting. We have helped with deploying Trac, SVN, CVS, LVS, mailman, Xen, and a raft of other applications and services to help keep our hosted partners focused on one thing: developing great open source software. All of the things we do are just distractions to these projects and something that we’ve been able to commoditize because we do it on a (relatively) big scale. The OSL’s infrastructure team (Corey, Eric, Matt, Trevor, and Michael) all take pride in working really hard to go above and beyond the call of being “just another co-lo.” These guys really care about this stuff and it shows every day in the work that they do.
NF: What did you start with in terms of infrastructure, and where is it today?
Kveton: We started with literally two Shuttle XPC boxes and the 72G of RAID5 disk for mirroring. That was about three years ago now. Today we host 120+ servers for about 40 projects. This is housed in about four full racks plus a bunch of bread racks. Yes, those original Shuttle boxes are still alive and kicking. 🙂 If you check out our machine room webcam, you can see the latest addition; 30 plus racks of space for the OSL. That is slated to be completed by June of this year.
NF: How is it that open source projects have started with free or donated hosting and bandwidth support, and are now able to actually pay for it?
Kveton: The gist of it is, as these projects get more successful, they are able to actually fund their hosting, usually through community contributions, other times through corporate donations. Some projects cannot cover their costs but are so integral to Linux or the desktop that we consider them special cases.
NF: Are you a sought-after hosting provider in these circles?
Kveton: We provide pretty basic hosting needs but really do our best to provide the best that we can with what we have. Our guys work their hearts out to help support the open source community and I think it’s reflected in the fact that we’ve never had a partner leave us.
NF: Do you have to turn projects away?
Kveton: The only projects we’ve had to turn away were those that don’t fit our focus on community-driven, open source projects. We have had several vendors come to us asking to host their community sites (the .org to their .com piece) and we just can’t with our arrangement with the University. Plus, its not really sticking with our core mission of supporting community open source projects. We’ve never actually advertised our hosting efforts. Its been completely, well, open source really. People just show up at our door referred to us by somebody else who is already hosted here.
NF: Is the OSUOSL getting adequate support back from the open source communities that rely on it?
Kveton: The open source communities themselves do the best they can to help us out. They scrap together machines, memory, and hard drives to get us the machines that they need to be hosted here. The demand for our services is increasing, and we’re hoping that with the new data center online, we can attract some of the industry to step up and donate equipment and money to help support the open source community hosted here even more.
We truly believe that these open source communities are where the true innovation is happening today in terms of collaborative technologies. These communities are creating software that is leading to more than just more software to help geeks write more software. They are laying the groundwork and developing methodologies for people, any people, to collaborate on a global scale without any boundaries.