"Gitlab has some growing left to do, especially on the SaaS part."

Some people (say, big companies) think exactly the opposite because they are after an on-prem solution, and for that Gitlab kicks Github's ass.

GitHub offers on-prem as well. Of course it's not open source and not potentially free like GitLab https://enterprise.github.com/features

$250/user/year, blocks of 10, min 10

Ouch. If you run your own servers already it's financially worth hosting your own gitlab at the minimum price point.

I expect it will be the big orgs who don't move due to integration costs.

Let's be honest, does anyone pay full price beyond 10 users? If there's anything I've learned it's that the word "enterprise" means "negotiable." I suspect they put that out there so they have a place to start without the costly interaction of emails coming in saying "how much for x people."

When their very first price point leaves it cheaper for me to do in house I'm not wasting time negotiating.

I like Github's UI slightly better (though it's been about a year since I've tried Gitlab)... the recent db issue has me concerned though.

That said, I think it's always worth considering using a tool because of their options for the open-source community... github, travis-ci etc have changed the way a lot of people work. GitLab might be what I reach for for self-hosting, but more likely to try the community/open edition first.

Using VSO (VisualStudio.com) for hosting at work, and some of the integration parts are very nice... Though no idea what a private TFS instance costs.

I use Gitlab and the DB issue only affected me in regards to CI, and I believe this is true to anyone that was affected at all.

They did a manual Backup before meddling with stuff so they only lost data from a really small timeframe.

CI downtime, on the other hand, is troublesome. I love using Gitlab CI and it didn't take me more than a few minutes to manually deploy what I had to, but I decided I need most of my CI to be controlled by me.

The end result is a simple refactor of my test and deploy specs to their own rake tasks (Could be Bash scriots or whatever) and I'll have Gitlab run a simple gle command, if it goes down I run the command manually...

Now some eventual downtime is not such a deal breaker anymore, and I continue having private repos and CI (Awesome CI) for even the smallest (And private) projects I work on... Not a bad deal if I do say so myself.

Why? Github enterprise offers exactly that. An "On premises" Github.

Gitlab Enterprise pricing starts at $39/user whereas Github Enterprise pricing starts at $250/user so from that perspective Gitlab kicks ass.

where did you see GHE asks for $250/user? seems the most expensive package is $21 [1].

[1] https://github.com/pricing

$21/user/month billed annually in packs of 10 (what your source says) would be $252/user/year in $2520 increments. It's actually (per the GHE features page) $2500/year for each 10 user pack though, which is $250/user/year.

GitLab Enterprise Starter is $39/user/year, and GitLab Enterprise Premium is $199/user/year.

From this page: https://enterprise.github.com/features

$21 is per-month. Per-year that works out to $250/user. If we're going by the month, GitLab's pricing is $3.25/user.

Having never used GitLab, what do you mean by growing from a SaaS perspective?

Performance of repos on GitLab.com https://about.gitlab.com/gitlab-com/

We hear you and agree, GitLab.com is too slow. We're working on multiple things to improve it https://gitlab.com/gitlab-com/infrastructure/issues/947 GitLab 9.1 will have lots of performance improvements for database queries. The GItaly team is working hard to move git commands to the fileservers to reduce latency, that will happen in about a month from now.

Idea: put the performance plots from tickets on https://arewefastyet.gitlab.com

Mozilla does this to great effect:

http://www.arewefastyet.com/

https://areweslimyet.com/

http://www.arewewebyet.org/

Great idea! We've just forwarded http://arewefastyet.gitlab.com/ to http://monitor.gitlab.net/dashboard/db/fleet-overview

This doesn't show all the graphs we need to show like 99th percentile request latency but that will come when we add Unicorn support to Prometheus.

This is the reason we love you guys, clear communication and caring for what people have to say. Keep doing that and we'll keep loving you. :D

FYI if I visit the new subdomain over HTTPS I get a cert error.

It only works on http I'm afraid.

Awesome! :D

As a thought for the performance diagnosis part of things, if your network switches have the option to enable NetFlow (also known as IPFIX), it can be extremely useful.

It does need to be paired with something that can capture and display the NetFlow data well. But with that, you can literally see what data is going where in your network, accurate to the byte. And with a good visualisation solution, you have the data historically too, so you can (easily) match things up time wise.

eg a developer writes a bad query which pulls 1/2 the database over the network to the front end server for filtering results there, instead of filtering them in the database. You can see that on the network fairly easily, and let the developer self-educate (if given access to the tools). ;)

As an aside, the gold standard used to be a commercial product called "NetFlow Tracker". It was amazing (fond memories), but the company behind it (Fluke Networks) didn't seem to know anything about software sales. So, it's now discontinued. :(

Hopefully there's a modern version of that somewhere. :)

We're working very hard on making GitLab.com much faster. Right now, it's measurably (and more importantly: measured [0]) faster than it was a month ago.

If you would like to follow progress you can dive into it all here [1]. One of the spearpoints is Gitaly, which should significantly improve git access.

[0]: http://monitor.gitlab.net/dashboard/db/fleet-overview?from=n...

[1]: https://gitlab.com/gitlab-com/infrastructure/issues/947