We received a lot of visitors on our booth throughout the day, and it was great to see so much enthusiasm around what our technology can bring in terms of scalability. Thanks to all of you who visited us and shared your stories with us.
Despite the craziness, we also took the time to attend the sessions. Here’s a quick review of this fourth edition.
Morning session:
Building scalable systems: What you can Learn about Erlang:
Mickaël Rémond described the strengths in terms of scalability that Erlang brings. Erlang VM provides excellent vertical scaling with its efficient concurrency management as well as a built in clustering management, which ensure easy horizontal scaling. Erlang agent model simplifies developer’s life and code readability. Makes us want to know more about it.
Why distribute graphs processing:
Vasia Kalavri talked about the quick evolution in the graph processing theory during the last decade. Apache Gelly for Flink, the lib she collaborated to, simplifies the development of graph analysis applications. Regarding the complexity of the theory, a lib to simplify usage seems more that useful.
Statefull containers:
Sandeepan Banerjee, former Google head of data, introduced the problematics related to statefull containers (containers + data). ClusterHQ offers Flocker, an open-source container data volume manager for Dockerized application which can help.
CERN collider IT:
Oliver Keeble from CERN described the complex IT and politics to deal with in order to manage and compute results from peta bytes of data generated by a single particles collision in a few hours: he described it as doing “Collision as a service”. Impressive!
Afternoon sessions:
After so much emotion, lunch was very welcome, and we have to admit that the quality of the food was really above average for a tech conference. Thanks guys!
During the break, our very own Audrey Neveu had the opportunity to perform a live demo at our booth on how to turn every API into a real-time stream of updates with no server coding, using the International Space Station position API as an example. Read more on her blogpost here!
Lightning talks:
Afternoon starts with the lightning talks; speakers presented us four different topics with only 5 minutes each:
- How to monitor at scale.
- Converge replicated data types.
- The most interesting uses of blockchains other than crypto-currencies (like proof of existence for documents).
- The usage of “mono repo” and “many repo” approaches with git.
This lightning format was too much information in such short time; even if subjects remained interesting, it left us a bit hungry for more.
IPFS:
Juan Benet introduced IPFS for “InterPlanetary File System”. A new peer-to-peer hypermedia protocol designed to solve drawbacks inherent to the world wide web (inefficiency, no offline availability, …). Ambitious and promising project.
The Internet archive:
Greg Lindahl described the Internet Archive ecosystem, the tremendous growth of the amount of data it deals with and the difficulty to build a search engine on tenth of petabytes of archived data.
Scaling Learning on Apache Spark:
Sean Owen described how to use Spark for machine learning the best way. ALS (Alternating-Least-Square) algorithm and its optimizations were discussed in details.
Transaction model – Goals:
Spencer Kimball described the constraints and solutions for being able to handle distributed transactions. It shows how distribution is not always improving performance, but can ease elasticity.
Cloud data saving:
Gleb Budman told us the story of the Backblaze “storage pod” genesis. How they used the agile methodology to develop their hardware and how fun this was. Definitely a must see!
Conclusion:
Sadly most of us had to catch an early train back to the Alps, and the remaining ones were busy greeting the visitors at our booth, so we were not able to attend the final session of talks and the closing keynote. But hearing the clapping and sheering from outside the room at end, it sure was a big hit.
This was a really fun day for us, where we learned a lot, met a ton of very interesting people, and interested most of them (we hope!). The organization was awesome and the speakers very high level, making DotScale a must-see tech conference in Europe.
A big thank you to the dot team for a job well done, we will definitely see you next year!
PS:
You have missed the event and want to catch up? The first videos are now online on The dot Post!