2015 is dead, long live 2016
A pretty serious list of ridiculous/amazing shit happened in 2015. My face was on a fucking pillar. I got to meet KEITH BOSTIC. A bunch of my favorite engineers got dressed up in cheetah kigurumi pajamas and ran around playing booth babes. And that’s just a three day sampling of mischief.
There are at least three key things about 2015 that I feel like I am always going to associate with this year, so that’s what I want to write about. (I’ll do a boring second post later with links to all the talks, articles etc for the year.)
The three things that stood out for me this year were:
- Conference Overload
- MongoDB + RocksDB (aka “mongo grows up”)
- Leaving Parse.
The Year of Conference Insanity
In 2015 I gave talks at 21 conferences, which was … excessive.
This wasn’t really intentional. But I was bored and restless and unhappy a lot and had many erratically available spare cycles. Being a manager at a big company makes this depressing Swiss cheese out of your calendar. You can’t really carve out the heads-down time you would need in order to do challenging technical work, but it’s easy to go on lots of short trips to conferences. And travel is nice. So I just kind of said “yes” to everything.
I really appreciate all the brilliant people and organizers who welcomed me to their events this year, and all the many random, unexpectedly lovely connections I made there.
Next year will be very different.
MongoDB + RocksDB
When I started running MongoDB in 2012, we were on version 2.0 and it had a single lock. Yep, just the one! This was a SUPER fun way to try and build a sophisticated multi-tenant platform.
Mongo was rolling out performance improvements with every release. But shit actually got real when they put the storage engine API on the roadmap and acquired WiredTiger. The Facebook RocksDB team enthusiastically pitched in too, providing feedback on the API design and ultimately delivering a full-fledged implementation of MongoDB with RocksDB storage engine.
Parse spent the past 1.5 years being the alpha customer for MongoRocks. We developed a load testing framework for replaying production workloads, evaluated TokuMX and WiredTiger as well as Rocks, worked with the RocksDB team (aka Igor Canadi) to iron the kinks out of the MongoRocks implementation and figure out how to run and monitor the damn thing, and then got it rolled it out to 100% of our production replica sets in under a year. (Worth it? with Rocks we used 1/10th the storage space and writes were 50-200x faster compared to mmap.)
I’ve said this before but I really believe it: 2015 is the year that MongoDB grew up and became a “real database”. This was Mongo’s InnoDB moment. Remember how much shit everyone used to give MySQL back in the MyISAM days? Hey guess what, software actually can get better! This is one reason it has been a special delight getting to work with Mark Callaghan. Mark was one of those engineers who turned MySQL from a punch line into a “real database” which powers much of the internet. So it was kind of like having this guy around.
I loved having a front row seat to mongo’s awkward adolescent phase. I loved getting to play a small role in promoting storage engine diversity. With Percona now offering enterprise support for both MyRocks and MongoRocks, I think the project is in good hands, and I’m really really fucking proud of this.
Leaving Parse
I said goodbye to Parse in early September. This was one of those emotionally overwhelming life changes that somehow manages to feel both too soon and way, way overdue. I spent 3.5 years working on Parse — one year pre-acquisition, 2.5 years at Facebook.
I like to think of myself as this crusty backend engineer who sighs wearily and doesn’t give a shit about product, but the fact is I fell in love with Parse on the very first day. I am so, so proud of what we built. We built a thing that people really cared about. We empowered so many developers to build and create and make whole new businesses on our platform.
There is nothing as heady as getting to pour yourself into a product that people are passionate about, tackle problems that are hard and might actually be impossible, with a team that blows your socks off with their brilliance and joy and hilariousness on a daily basis.
So yeah, it was hard to let go. But it was time. The 3-4 hour commute was fucking killing me, and it had been a long, *long* time since I felt like I was learning new things or pushing my boundaries as an engineer or a leader. I need a certain amount of chaos and panic in my life. Being comfortable for too long just makes me miserable.
Parse is going to be a hard act to follow. Thanks for raising the bar, motherfuckers.
In summary
2015 was ridiculous and awesome, and I want 2016 to be ridiculous and awesome but in some radical and fantastically new ways. I have many hopes and intentions for 2016. but I’ll save that for another time.
Oh but my single greatest achievement of 2015? is *clearly* this tattoo.
12 months, 11 appts, about 30 hours of inking. thank you micah riot.
