12 minutes reading | Também disponível em 🇧🇷
TL;DR I’ve been at OSCON this year (thanks to Locaweb) and this post will summarize my impressions about the conference. I spoke at Locaweb Tech-Talks about it and you can find my slides here (in portuguese).
OSCON (Open Source Conference) is one of the most important open source conferences of our time and happened in Portland this year as usual. Portland is a great city, the one I liked the most until now, it doesn’t have high car traffic, MAX (the train) took us for most of the places we wanted to know, nice people all over the place, great convention center and a lot of other benefits.
The conference started at 9:00 am of July 20, 2015 with the first day of tutorials. In the first day I followed the Kubernetes tutorial and it was cool. I had the chance to talk with Google guys about it and try Google Computing Engine, which was very nice.
The second day was filled with more tutorials and I chose RUST and Cassandra.
Was a good tutorial with a lot of exercises and good tips of the language. I really liked it and I’ll probably give Rust a try on the next months.
I was expecting a different thing from this one. It was based on how to use cassandra, create tables and so on and I was expecting tips on how to create and scale it and maybe a little bit about the caveats we may found on the way. But it wasn’t bad at all, I’ve learned a little bit on how to use it and as a gift I tried a little bit of spark as well.
At the end of this day the Expo hall opened for everyone, cool things there (including some T-shirts… T-shirts everywhere).
Let the conference begin! :)
It started with the keynotes at 9:00 and I would like to highlight one of them. All keynotes are avaiable on youtube btw. :)
Facebook open source at scale (Keynote)
I like to understand the open source culture behind these big companies. It is really interesting to know that there are only 2 people to handle all the facebook open source process shocked me a litte bit. They have a great open source review and publishing process and all the softwares they release are running in their production environment. So, why do they do open source software? Because it’s good for the company, helps them to make better software and bring smart people to work with them.
And then the multiple sessions started, a lot of great subjects in the agenda and you have to choose one (!!!).
Roller Coaster Tycoon with genetic algorithms and Go
The speaker Kevin Burke spoke about his hobbie of playing with roller coaster tycoon game maps with Go and Genetic Algorithms in order to create de best roller coaster ever! It is not there already but the presention was really cool to follow.
You type “google.com” into your browser bar and press enter: What happens next?"
The speaker for this talk is Graeme Mathieson. I was planning to spoke about it before de last edition of DevInSampa (2012? Btw, it could happen again…) but didn’t had the guts to make it yet and there it is, someone made it before me. The pace of the presentation was not good but the content was cool (it was his first presentation). At least it made me think more about this topic and maybe I’ll try to speak about it soon.
After the lunch I tried to attend to Evolution of information security threats presentation but the room was full, so decided to go to Matrix guys presentation.
Fixing the fragmentation problem for real-time communications
Matrix’s is really interesting, the main idea is to make a protocol for messages, just like the email protocols do for our email, we may use our client and store our own email but it is not possible with today’s messaging like hangouts, whatsapp, telegram, skype and so on. They had a live demo piloting a quadricopter via chat, it was really cool.
High adventures in sniffing my own metadata
The speaker is Josh Deprez from Google Australia. He described his work using a Go program (home made) to analize his own internet traffic and which information he got on it. The presentation itself was cool to follow but the discoveries wasn’t so much interesting. :P
Building the $9 computer, or how I learned to stop worrying and love embedded Linux
This is the C.H.I.P project owner Dave Rauchwerk speaking about the story behind the project. This guy is hilarious and a real hacker, he made a gif camera just for the lulz and tried a product with it! Unfortunately it was too expensive, to be viable for selling, so he decided to create C.H.I.P to make the camera cheaper and C.H.I.P became the product by itself.
Continuous delivery and large microservice architectures: Reflections on Ioncannon
This last one was really interesting Kevin Scaldeferri from New Relic spoke about how they grown from a big and monolythic Rails application to 250 microservices. In short, it was a mess in the beginning, a lot of techlologies and deploy tools until they organized and found a standard for the language and deploy tools, it is really better now, but the old Rails app is still there.
A new day with an typical American breakfast at Denny’s and starting with keynotes at 9:00 AM again.
Making architecture matter (Keynote)
Martin Fowler spoke about architecture in general, there was some good quotes. One of the arguments that really make sense is that the today’s “Architect” is that guy that don’t write code (!?). How can a professional that doesn’t write code talk about how to write code? Another good point was about the user perspective about architecture it can’t see it anywhere but still need to pay for it! We need to argue using economic principles in order to be able to create a good architecture for our softwares because economics always win.
We have a lot of examples of products that were the market leader for their segment and it could not evolve anymore to the point that it was swallowed by another competitor. This is a sign of bad architecture.
You code like a sysadm
hm… This talk was not so good but it left a message for everyone: “Bad code is still better than no code at all”
And obviously: Good code is better than bad code.
So, instead of not doing something because you think your code is not good enough, do it and improve as soon as possible.
When Performance Counts: C++ in the 21st Century
The speaker threw some good arguments that haven’t came to my mind until this day: “Why is C++ getting more attention these days?”, we have C++11 being created and more people talking about it. In his words, the world was not in need for performance anymore, we had big computers and most of time the other languages (java, in his examples) are doing a good job to create what is needed without the need to manage memory by hand. But today we have some other hardwares that need this performance back, like tablets, mobile devices, cloud applications that pays based on cpu time and so on. Good catch man! :)
Creating an open source office: Lessons from Twitter
I made a bad choice based on the presentation title and changed rooms as fast as I could to follow this one.
Some notes on the presentation:
- Ensure projects have owners or give it to the community: this is the twitter example of Twitter bootstrap
- Always say Thank you for your project collaborators
- Lessons from mesos
- Apple uses mesos for Siri
- Open source foundations force good pactices on you (incubated aurora)
- Lessons for twencache
- Keep track with the upstream if you fork something
- Share your code with some peers to gauge interest
- open source project with collaborators is better
A good lesson that I got from this presentation is that community doesn’t come for free. Companies like Twitter have their own community managers, conference speakers, conference organizers, documentation writers and their developers to make everyone engaged with the project. Throw something on github and expect collaboration is not the best idea for OSS.
Your employess may become open source celebrities sometimes, and if they became popular they may leave the company. It is not so good for the company but the same popular projects may attact new talents as well. :)
Troll’s are everywhere and youy will always hear that your project is stupid and useless. The best way to deal with it is to take it as a compliment because people care enough to say it.
Mesos the OS for your datacenter
Mesos has a high level API to do things and you don’t need to care about the resources. It provide an algorith called DRF for a fair allocation of resources.
There are several built frameworks for mesos like Marathon, Aurora, Jenkins and some more. It’s easy to build your own framework (netflix, twitter, apple and two sigma have it). A good reason to build your own is to deal with some exotic workload, this default frameworks are built for small tasks.
For me, mesos is the future and I’ll definitely be looking for it.
This was the last day of the conference and it started with 3 keynotes. After the keynotes there was two sessions more and the conference ending at the main ballroom.
Hacking Conway’s Lay (Keynote)
organizations which design systems … are constrained to produce designs which are copies of the communication structures of these organizations —M. Conway
When rebuilding a system it’s really important to think about your team because it will be part of its architecture like conway’s law says. So when a decision is made to build a new system without changing the way people work you will face the same problems the other system had but with a different code.
Doesn’t matter if you’re using microservices but your team is still monolithic.
To solve technical problems, you often have to go outside the technical domain. Raffi Krikorian
Good points that ended the presentation:
- It’s not all about the code!
- Small teams can do big things.
- Full stack teams keeps your architecture flexible.
- Don’t fight Conway’s Law, work with it.
A lot of good lessons learnt from this one. :)
Build a botnet?
The speaker (Bryan Smith) did a great (and funny :P) presentation around the topic. And some things are really scary when speaking about botnets. The most scary thing for me was its price, you can use a botnet for an hour spending only $5… it is really low and probably will get a service out of business for that time. There are some softwares avalable if you want to build your own botnet these days (accorting to Bryan, if you want to build a botnet you’re probably a scumbag… LOL), here are some of them:
- kreios C2
- Andromeda V2.6 - leaked
- Kaiten IRC bot
- SDBot - GPL
- DSNX Bots
- Q8 Bots
I’ll not post links here but you can google them if you really want it. :P
The most funny part of the presentation was its end when the speaker laptop got rebooted when trying to bring the botnet up. LOL
Freedom and responsibility @Netflix: Centralized team in a decentralized world
One of the team roles is to increase the rate of innovation and reduce the impact of change by providing good tools for the other teams. Their clients are the other teams engineers, it’s not a simple internal system it’s a product!
Netflix has a great culture and each team as a subset of this culture.
Netflix has a lot of good tools but the engineers may choose to not use it and build their own in case it doesn’t fix their problems. For this reason there are 4 deploy tools running on Netflix actually.
They have no Ops team, you build it, you support it! And in order to remove part of the constant interruptions, the team uses a “responder rotation” when each time (day, week, month?) a new engineer will be in charge of solving problems and helping the other teams.
They (like many of us) don’t like ticket systems, but they use it anyway (like many of us?). For this case the tip is: Make creating a ticket easy, minimize fraction by making it as easy and fast as possible.
Hackatons are made between the team in order to keep the innovation process flowing and making space for innovation. Another good thing to think about is to always hire beyond your needs, it makes more space for innovation (not all people working on day to day tasks) and it will be really helpful when the project is on fire and the team needs more intensity on it.
Communication is really important for the teams and they do not set a limitation for important communications like framework versions and DC migrations. For this task they use hipchat, email, banners and everything that may call the engineers’ attention.
Situation normal, everything must change
This was the last keynote and maybe the best one, and I highly recommend you to watch it.
Simon Wardley used the fact that OSCON is moving to speak about situation awareness and what could happen when you’re not prepared for a new thing.
His arguments are that we’re using the wrong maps to try to understand our problems. In business we use SWOT for example, which is really bad to understand to understand the whole picture, it was really funny to see his comparisons with the battle of the 300 using a SWAT map instead of land map. LOL
The speaker shows some maps he uses for his business, and it looks really better than just a SWOT. :)
Again, I really recommend you watch this talk. I liked it. :)
The best conference I ever attended. Everything was great: hotel, Portland Conference Center, the city in general, the food, the conference content and so on. It definitely worth the 14 hours plane travel to be there.
I would live in a city like Portland, it is one of the best cities I ever visited. I’m sad to know that the next OSCON will happen in Austin instead of Portland. :(