The very first time we saw GitHub demonstrated, it was clear to us that it was important to the Ruby on Rails community, and something that we wanted to be associated with. In fact, we felt so strongly that we immediately offered Tom, Chris and P.J. free deployment on Engine Yard in exchange for free accounts for our customers and developers.
Why did we feel so strongly? I’d like to explain this on the record as many of our customers will no doubt wonder “Wow, that was a pretty good deal?” :-)
My number one reason was that I felt that distributed version control was critical to keeping the open source community centered around the Ruby on Rails cutting edge. The social implications of Git’s fully distributed nature ensure that open source developers must listen to their community or risk losing control of their project to a fork. This is why we moved Rubinius and Merb to Git so early in the game!
When Rails itself announced the move to GitHub, we were overjoyed and entirely supportive. We then moved GitHub from a small slice environment to a larger cluster as we felt it critically important that GitHub scale smoothly with the substantial increase in traffic that the Rails source code would bring. That transition went so smoothly that an impromptu party developed to celebrate its success; it was quite an evening!
Shortly thereafter, as often happens when a site reaches a certain level of success, things began to slow down. We identified the bottlenecks and supported GitHub and the community by making patches to ssh to allow key lookup in MySQL rather than a text file. That remains, to this day, one of the finest examples of Engine Yard support and it makes me extremely proud just thinking of it.
The Ruby and Rails communities have wholeheartedly agreed with us and endorsed GitHub. In fact, the transition from Subversion to Git has been one of the most aggressive technology transitions that I’ve ever seen! This made the ever increasing resources required to deliver GitHub a fair trade for Engine Yard for a number of reasons:
- More GitHub traffic meant more exposure for Engine Yard
- Git allowed the Ruby community to be more agile, and more GitHub adoption meant a more robust Ruby community
Recently, however, GitHub has begun to grow very rapidly in non-Ruby communities. This is wonderful for GitHub, but made adding new resources on a free basis far less attractive to Engine Yard. GitHub offers the largest free storage quota among the big SCM hosters, and we came to the conclusion that we didn’t want to subsidize that quota for non-Ruby developers.
We discussed options extensively with GitHub and made our best attempt to offer an easy ramp from free to paid. However, in the end, the best arrangement for them was moving.
We use GitHub each and every day for every project that we support. We worked incredibly hard for them since the beginning, and placed a substantial amount of resources at their disposal, in hardware, bandwidth, and support. Like all partnerships, we had our occasional disagreements, but I believe there were never hard feelings on either side. Accordingly, we are currently working directly with their next provider to maintain smooth service as they cut over.
For Private Cloud customers, nothing will change: our classic, fractional and dedicated offerings will continue to include a source code management account (Github or Beanstalk) as part of their service. For existing Engine Yard Cloud customers, we will continue to provide free Github through the end of this year. For new Engine Yard Cloud customers, there will no longer be a free Github account included in the service.
We’re confident that GitHub will continue its tremendous success in enabling the kind of collaboration that forms the heart of the open source software movement. We wish them the best and will continue to think of them as an important part of our future.
Loading ...
Watch a Live Demo of Engine Yard AppCloud
The Engine Yard Newsletter
I would love to hear GitHub's side of this story as well. The impression I am left with after reading this article is that:
1) GitHub received free hosting, bandwidth, and technical support (all tangible expensive items), in exchange for free GitHub accounts (non-tangible and cost free) for more than a year.
2) When the hosting, bandwith, support burden grew to a size that it was no longer in EngineYards interest to support a much broader non-ruby community for free, GitHub decided to pull the plug and go with someone else.
Is this accurate?
If so, this move reflects badly on GitHub and demonstrates no loyalty to those that provided the foundation they were built on. I think they need to make a public response to their paying customers.
PS – GitHub guys. I am no-tea-totaller, but you need to layoff the constant stream of alcohol related events, and posts on twitter, etc. Paying customers don't react well seeing party boys whooping it up when the service continues to suffer from frequent outages, slow queues, occasional inability to push/pull, frequent breakage after code pushes, etc. You need to be seen as serious about the product you are delivering as your customers are about theirs, and behave as though committed to it 110%. Please remember that other git hosting solutions are rising up and the opportunity cost of switching git providers is low due to Git's distributed nature.
I reluctantly have to agree with this comment. Although I haven't experienced the service problems on the level that Curious has, the Github operators seem to almost be actively pursuing a poor public image.
I was very disappointed to have wasted my time at the last RailsConf going to their session. I got absolutely nothing out of it and all they did was talk about themselves, not their service. And it does seem awfully ungrateful to just ditch EngineYard after the incredibly costly and superb resources they've provided for free for so long.
I think the github team's been struck with the Rock Star stick and feel motivated to propel the bad-boy image by sponsoring things like Drink Up Thursdays and awarding bottles of bourbon.
Right, awarding a good bottle of whiskey sure propels a "bad-boy image". They are right up there with Tommy Lee.
Getting involved in developer communities, whether via social networking online or in person, is part of the GitHub guys' jobs. Nobody can be productive at a keyboard twenty-four hours a day, seven days a week. Is it a crime for them to spend a couple evenings a month buying drinks for their customers and potential customers, and finding out how to better serve those customers?
I'm a paying GitHub customer, and I think they're 110% committed while also being shrewd networkers. If you went out and had a beer with them, you could find out exactly why the site is slow and what they're doing about it. They're trying to do a lot with a very small team, and they're doing a damn good job.
The next time you get up on your high horse to chastise someone for not being "110% committed", put yourself in that position. Do you have a life? Do you do anything other than work? Do your activities outside of work directly benefit your company? Get real.
Its teetotaller, not tea-totaller. And you're more off-topic than anything else.
"Curious" seems to have a largish axe to grind. Why not take the article at face value instead: Github grew beyond hosting Ruby projects so Engine Yard didn't think sponsoring them (in the old way) was a good idea anymore. Their prerogative. Case closed.
So why didn't GitHub convert to paying customer? Did they get such a better deal elsewhere? That would be nice to know.
So you're happy to subsidise Ruby developers, but you aren't happy to subsidise Open Source developers in general? That seems like a pretty poor attitude. I have no problem at all with GitHub outgrowing a free hosting plan, but explaining it in those terms leaves a bad impression.
Simon: I think what's missing in your thought process is that EY isn't subsidizing Ruby developers. It's a services exchange. EY trades hosting for advertising. Since EY deals almost exclusively with hosting Ruby apps, they are interested in advertising to Ruby developers.
As Github's audience starts to expand into other languages, the effectiveness of their advertising decreases, reducing the value EY gets from this trade. All the while, Github's load increases and so Engine Yard's costs rise.
I think Tom was trying to explain that in this article. Engine Yard's costs were increasing while their benefits were decreasing. There's no subsidizing here, it's a trade. When the trade imbalance becomes too great, you need to revisit your agreement.
That was very well explained Kyle. Its not about not-caring about open source developers. EY's target customers are Rails/Merb developers, and would get more leads from them through the adverts.
I too feel, Github could have just converted to a paid account. Would love to hear from Github guys the reason behind this switch, it could be a valid one. We don't know.
For me, it's all about the way this was communicated. I think Engine Yard are 100% within their rights to not want to continue subsidising GitHub (a profitable company) in this way. I think the business decision that it doesn't make as much sense now it's not just Ruby is a perfectly reasonable one. But when you actually go right out and say it in the way it's said in this post, it's going to leave a bad taste in the mouth for other open source developers.
+1. "We came to the conclusion that we didn’t want to subsidize that quota" is fine… "For non-ruby developers" just makes it sound like a couple of boys in a treehoue with "no girls allowed" signs.
Sounds like a fairly normal commercial decision to me, not a poor attitude. They are in the Ruby business, so they foster the Ruby community. Fair enough. GitHub's growth coming from outside the Ruby community and taking EY resources? Rational decision, I'm sure it wasn't rushed or made without consideration.
Thanks for the kind words Dom. We do our best, and you're absolutely correct: this move was made with a lot of planning and thought by both GitHub and Engine Yard. We're confident it'll all work out for
the best :)
I have to agree with Simon. I am very grateful for you providing this free hosting for Github, I'm sure without it it these costs would have slowed Github's growth and thus made it hard for open source developers to have their code out in the open.
But in the end, not wanting to serve the open source community as a whole is a bad way to spin things. Ruby does not exist in an open source vacuum. Ruby, rails, etc borrows elements from projects written in other languages, and other languages borrow from Ruby. It's how languages grow, and its how open source works.
In any case, thank you for what you've done and I'm sorry it has to end, but I wish you'd reexamine your biases.
Tom & all at EngineYard: thanks for helping GitHub get off the ground. It's a brilliant idea and kudos to you for realizing that from the get-go. Transitions are never easy, but sounds like you guys are handling it like pros.
Thanks Luke! We're doing our best, as I'm sure are they :)
The question that's really left open here is, "why would Github choose to go elsewhere instead of pay Engineyard?" If EY really wanted to keep them on, but couldn't support it for free, then surely they could have discounted their services to the point that Github couldn't have found a better deal elsewhere. So, one way or another someone really wanted to split up the partnership… the question is just who?
> "why would Github choose to go elsewhere instead of pay Engineyard?"
Well, presumably they felt they could get more service for less money than EY were offering, elsewhere. Why should it be anything more complicated than that?
Engineyard is a great choice for some companies, but it's extremely expensive. I'd imagine that if GitHub switched to a paid plan it would cost around $20K per month. Github is probably able to do much better with a company that offers better rates on hardware and perhaps a bit less support.
Don't listen to Sarah
Why hasn't GitHub announced this (fairly major) change to their customers on their blog? Major hosting transitions are never easy, and rarely come to pass without customer impact. You would think that a heads up to the community is certainly in order.
As for Engine Yard, I don't think they have to answer to anyone regarding their commitment to open source, and especially their focus on the Ruby and RoR communities. They are arguably the most influential company in this space. Lets see, Ruby 1.8.x maintainers, JRuby, Rubinious, Rails 3, Merb all funded out of their own pockets. And Yehuda Katz too! Do some of you expect them to support *all* of the open source world out of pocket? Thanks @ey for what you have given all of us.
I for one would like to hear the back story from the GitHub guys as well regarding the effect of this transition (if any) on their business. That is unless they're too busy sharing a bottle of 23 y.o. Pappy van Winkle [ http://bit.ly/x7giB ] with the 'functional alcoholics' [ http://bit.ly/3LJ507 ] . ;-)
I want to hear Github's side because I think I'm missing something. If based solely from this side of the story, it seems like Github certainly would be better off staying with Engine Yard as a paying customer: EY's support was good, they delivered beyond expectation and they certainly brought a certain amount of success to Github. But in the end, Github still decided to pack and leave rather than pay and stay.
There must be some compelling reason for Github to make such a weird decision. The question is what?
Perhaps they want to manage their own servers and software. At a certain scale that becomes cost-effective.
The key phrase is "I believe there were never hard feelings on either side".
I'm not convinced that that is 100% accurate.
Seems pretty straightforward to me. I've met some of the folks at Engine Yard and they are solid people. The folks at GitHub seem the same way. There's just a point where it no longer makes sense to keep giving away valuable services for free.
In case you missed it, GitHub has a blog post explaining (in great detail) their move to Rackspace.
http://github.com/blog/493-github-is-moving-to-ra...
Curious: No, not accurate. It was a trade of services for visibility, not a down payment on future business. That's the nature of an at-will agreement. They both did their part and grew because of it.
Curious' trolling aside, anyone who has watched Github for more than a month knows the level of engineering they do (erlang gitd, pages/jekyll, gist, FI). They're incredibly committed, and speaking as a customer, my value per subscription dollar is incredibly high. Given the one-of-a-kind scaling problems I'm sure they've encountered, I'm amazed it's been as responsive and stable as it has. We push and pulling constantly, and 99% of the time, it's snappy.Curious: No, not accurate. It was a trade of services for visibility, not a down payment on future business. Both parties grew because of it.
Curious' trolling aside, anyone who has watched Github for more than a month knows the level of engineering they do (erlang gitd, pages/jekyll, gist, FI). Given the one-of-a-kind scaling problems I'm sure they've encountered, I'm amazed it's been as responsive and stable as it has. We push and pulling constantly, and 99% of the time, it's snappy. They're incredibly committed, and speaking as a customer, my value per subscription dollar is incredibly high.
As for the social element, read the Web site: social code hosting. Some of us enjoy meeting in person, too — I'd sure attend a meetup in Seattle.
@Dave,
Lot more than $20k, probably more like $50k. They are very expensive, and the problem is it doesn't really buy you much, the rolls royce of cloud computing is sortof a joke.