The 2018 Linux Conference Australia was held the University of Technology Sydney from 22-26 January 2018. I attended courtesy of eResearch.

Linux totally dominates supercomputers. As of November 2017, all 500 of the world's fastest supercomputers were running Linux. This is because most of the world's scientific software for generating or crunching research data is written to run on Linux systems because of its openness, ability to be customised, speed and robustness. The eResearch HPC runs the Centos 6.9 Linux distribution and some of our users run a linux "distro" themselves.

Like most large conferences there were many parallel sessions, so deciding what to attend was sometimes a difficult choice. The sessions I attended mostly reflected my interest in bioinformatics and system administration, as those topics have relevance to the HPC cluster administration and user support.

Of particular interest in the bioinformatics stream was; James Ferguson and Dr Martin Smith of the Garvan Institute and their nanopore sequencing pipeline, the data and code that Stemformatics and CSIRO's bioinformatics provide, Matt Tods keynote on the Open Source Malaria project and key requirements of lab notebooks (read some of their open Lab Notebooks).

In the system administration stream of particular interest was Terraform for building and versioning infrastructure safely, James Shubin's new configuration management tool Mgmt, using mod-security and fail2ban, and Thomas Schöbel-Theuer's MARS Long Distance Replication which replicates huge amounts of data across continents.

The talk on high performance science covered data formats, provenance and reproducible research. If your using git and don't mind a good command line tool for reproducible research then have a look at DataLad. This is build on top of git-annex and extends it with an intuitive command-line interface. ReproZip allows you to pack your research along with all necessary data files, libraries, environment variables and options. It's a python package similar to recipy which also attempts to help with the problem of reproducible research.

Mentioned was an interesting site binder, which turns your GitHub repo into a collection of interactive notebooks (consider security though). If you use pipelines for your data flow then here is a curated list of awesome pipeline toolkits, apparently inspired by a curated list of amazingly awesome open source sysadmin resources.

Other personally interesting talks were the MUON detector, and stories from The Register's "On Call" and, if you are an open source developer, an excellent talk on the problem of contributor overload.

LinuxConf 2018 Stats: 200 presentations selected from 400 submissions, 700 attendees from 20 countries, 140 MB/sec sustained bandwidth for 1.8 TB during the conference, 1321 unique network devices detected, and more than 2000 coffees served.

The next Linux Conf will be held from 21-25 January 2019 at the University of Canterbury, Christchurch, New Zealand. See the link here: LCA 2019, Canterbury NZ. Registration costs for students and hobbyists is always kept low and Linux Confs are always very inclusive as diverse groups of people are made welcome.

Mike Lake
eResearch