SCALE14x trip report

SCALE14x was held at Pasadena, Los Angeles this year from January 21-24 2016. I think its important to note that the venue changed from the Hilton LAX — this is a much bigger space, as the event is much bigger, and you’ll also notice that the expo hall has grown tremendously.

I had a talk in the MySQL track, and that was just one of over 180 talks. There were over 3,600 people attending, and it showed by the number of people coming by the MariaDB Corporation booth. I spent sometime there with Rod Allen, Max Mether, and Kurt Pastore, and the qualified leads we received were pretty high. Of course it didn’t hurt that we were also giving away a Sphero BB-8 Droid.

The MySQL track room was generally always full. We learned some interesting tidbits like Percona Server 5.7 would be GA in February 2016 (true!), the strong crowd at the MariaDB booth and quite a bit more. People are definitely interested in MySQL 5.7’s JSON functionality.

The highlight of my talk, The MySQL Server Ecosystem in 2016 was that it brought along quite a good discussion on Twitter. Its clear people are very interested in this and there is much opportunity for writing about this!

The Mark Shuttleworth keynote

But there were other SCALE14x highlights, like the keynote by Mark Shuttleworth. It was generally a very moving keynote, and here are a few bits that I took as notes:

  • Technology changes lives
  • Society evolves because it becomes possible to live differently
  • New software moves too fast for distributions (6 months is too long). Look at Github. Speed vs. integration/trust/maintenance (the work of a distro)
  • snapcraft 2.0 (learn more about your first snap): reduce the amount of work to package software. Install software together transactionally.

An overview of a next-gen filesystem

Another talk I found interesting was the talk about bitrot, and filesystems like btrfs and ZFS. Best to read the presentation, and the article that was referenced.

Scaling GlusterFS at Facebook

A talk by Facebook is usually quite full, and I was interested in how they were using GlusterFS and if anyone has managed to successfully run a database over it yet (no). This was a talk given by Richard Wareing who’s been at Facebook for over 5 years:

  • GB’s to many PBs, 100’s of millions of files. QPS (FOPs) is 10s of billions per day, namespace (volume), TBs to PBs and Bricks: 1000’s. Version 3.3.x is when they started and now they use 3.6.x (trail mainline closely)
  • Use cases: archival, backing data store for large scale applications, anything that doesn’t fit into other DBs
  • RAID6, controller is enterprise grade, storage is more consumer grade
  • Primarily using XFS, and are starting to use btrfs (about 20% of the fleet run on it)
  • closed source AntFarm, JD, and their IPv6 support (they removed IPv4 support). They have JSON Statistic dumps which they contributed upstream.
  • a good mantra, pragmatism over correctness

Some expo hall chatter

There was plenty to followup post-SCALE14x with many having questions about MariaDB Server, or wanting to buy services around it from MariaDB Corporation. I learned for example that Rackspace maintains their own IUS repository of packages they think their customers will find important to use. The idea behind it is that its Inline with Upstream Stable. Naturally you will find MariaDB Server as well as packages for all the engines like CONNECT.

I also learned that Stacki uses MariaDB Server for provisioning, as was evidenced by their github issue.

Its incredibly rewarding to note that pretty much everyone knew what MariaDB Server was. Its been a long journey (six years!) but it sure feels sweet. Ilan and his team put on a great SCALE so I can’t wait to be back again next year.