Industry leading domain registrar and web hosting company
59 million domains under management
700+ new tld extensions
Over 5 million sites hosted
Over 35,000 servers
Everything you need to have a web site and run a business online
Our passion is enabling small business owners
Target customer is Very Small Businesses (5 or fewer employees)
28 million small businesses in the US, 85% are VSB
200 million self-employed people outside the US
Focus on growing their business, not on running their web site
Huge business opportunity! We need to be able to scale quickly.
So why OpenStack? Recognized our systems & processes were dated
Needed to be more agile and flexible - OpenStack fills this gap
Truly self service cloud platform has enabled an explosion of use cases
This in turn increases agility, innovation, and efficient across the business
Joke about standing up POC during the product meeting
Real benefit is OSS community. Many people working on it, a lot of expertise
Free/financial/liberty, able to make changes to match our use cases
Want to show you the process for getting where we are today.
Start with the playground
All developers/engineers in the company have some personal quota
Help folks to recognize how to get started, understand and experience the benefits
This has been wildly popular, few hundred VM’s within a couple months
Turns out people really like to do stuff on their own, instead of opening tickets!
Next is dedicated formal projects for development teams
Really speeds up development, because teams can easily spin up new servers, etc.
Almost all our dev teams are using OpenStack in some way for dev processes.
As a result, over $2 million in savings last year just from dedicated dev/test servers we didn’t have to buy
Early adopters, teams that are already “cloudy” and are on the cutting edge
Maybe were already using AWS or some “roll your own” cloud thing
These guys are just chomping at the bit to get on the cloud
Makes them happy, but also helps to prove the cloud for production
And bubbles up areas to address for wider prod usage (affinity, maintenance, etc.)
~1000 cores
After that we did a new Hosting VPS/”cloud servers” product
This is in beta now, google “godaddy cloud servers beta”
First truly “customer facing” thing, with potential to grow really fast (1000 cores so far)
Rubber hits the road here, this is what really proves the cloud works
If it can handle this workload, it can handle anything
Finally, systematic and mandated on-boarding of all applications
This is what we are working on now, up to about 4000 cores across two regions
Meeting with individual teams to go over their architecture and how to make it cloud ready
This can be challenging for some apps that prefer the old way, Puppies vs. cattle
Expect to have most apps on OS by EOY
Then we’re on a consistent platform across everything, easier to manage & scale
This is where we are today, but want to take a trip back in time to before OpenStack…
To get a server: high amount of process and procedure
Average time from request to server ready for app owner: 3 weeks
Build checklist which created tickets to individual DC, networking, etc. teams to do their parts - 100+ steps!
Good intentions here… needed process to handle fast growth, but Ultimately not scalable
This was true for dev/test servers, too. On Vmware, but treated just like physical servers
Slows innovation and kills ideas, because there was such huge resistance to getting resources to POC idea
Imagine how excruciating that is, even just watching this calendar raises your blood pressure.
This situation leads to is … server hoarding
Choose your favorite metaphor for highly valued items
3+ weeks and a big hassle to get a server, no way am I giving back any that I already have.
Even if I’m not using them, I might need them later.
Can’t really blame people for this. It’s just economics.
A lot of effort to produce something, it will become very PRECIOUS, especially once you have it.
Leads to a lot of waste
Hardware utilization on non-OS hardware is only ~8% (based on CPU/ram)
That means over 90% overhead on DC space, power, cooling, network ports… all that stuff.
Now we can do about 3 minutes to get a server
Totally changes everything, easy to get a new one, no problem to throw it away (ephemeral)
Huge benefit to development teams immediately, really popular
Lots of little internal tools popped up
Give people tools and trust and they’ll do great things
They can focus on their app
Encourages more automation and orchestration
Manage their own infra without really having to think about the infra
Also have seen great hardware efficiency gains
Fundamentally people don’t want to be wasteful, they will be efficient when given appropriate tooling.
Can orchestrate infrastructure on the fly, allows you to only use what you need.
Hardware utilization on OpenStack now more than 60% (huge gain from 8%)
We still want to make this better
The the biggest benefit we see is the power of the community …
This is the real power. Over 3000 contributors and nearly 8000 actively participating in bug reports, code reviews, etc.
Everybody sharing experiences, war stories, best practice, help with problems
Operators group valuable for this, great way to be involved other than code
People generally are very open about sharing configs, patches, etc.
Enabled us to go from nothing to private cloud in just a few months with only 5 people
Image credits:
http://www.mattgriffin.com/2014/04/26/open-source-appreciation-day-draws-openstack-mysql-and-centos-faithful/
http://www.mattfischer.com/blog/?p=610
http://superuser.openstack.org/articles/superuser-weekend-reading-063ef69e-394d-4cc7-a11c-ffba9662dd44
http://www.booksprints.net/2013/02/open-stack-book-sprint-underway/
http://www.openstack.org/blog/2013/04/women-of-openstack-at-the-portland-summit/
https://twitter.com/amitry/status/574942441373069312
http://princessleia.com/journal/?p=8526
http://princessleia.com/journal/?p=9268
http://www.huffingtonpost.com/ben-lamson/how-to-successfully-crowd_b_4274202.html