Latest evolution of Linux IO stack, explained for database people Sunday 16:00 Devroom
Twitter: @the_hydrobiont Twitter: @dataegret LinkedIn: hydrobiont LinkedIn: dataegret Facebook: hydrobiont Facebook: dataegret Company website: www.dataegret.com Blog: blog.postgresql-consulting.com
My name is Ilya Kosmodemiansky and I am a Co-founder and a CEO of Data Egret. A firm providing a remote DBA support for PostgreSQL. Мy first encounter with databases was actually in bioinformatics, then I worked with Oracle, DB2 and, for many years now, with Postgres and I must say that I really enjoy being a part of this community.
In addition to helping with some of the organizational aspects at dedicated Postgres events, being a part of Code of Conduct committee, arranging community booths and giving talks and workshops, I am very lucky to work with most talented Postgres DBA whom I encourage to contribute to Postgres development and share their knowledge with wider community giving talks at events and through our blog.
In an open source, a community is more than just a group of people who use the technology or those who write it. The magic happens when the two come together and those who seek the improvement take the lead. This is when the technology really thrives.
PGconf.EU and FOSDEM are two events that Data Egret traditionally attends. I normally come as a speaker and some of my colleagues often join me to meet the community and support the events.
The community aspect of both conferences is truly amazing and I would highly recommend attending both to anyone who is working with PostgreSQL or considering using this technology. PostgreSQL is not just about database - it’s about people.
There are several reasons to this. First, the way Postgres evolved, it is now at the stage when it needs the diversity of database requirements that can only come from IT colleagues in different industries. In the early days, there was a very small group of developers who launched the project and grew it to the point when it could be nurtured further by a larger community. Due to the nature of such development the project acquired structure that, at times, hinders from addressing some architectural challenges. Hence, attracting new people who will have a fresh perspective on the project is essential for addressing these issues.
Second, to fully discover its full potential, project developers need to hear from as many clients from various industries as possible rather than focusing on finding solutions for big customers. One of the perks of open source is that this is actually possible, so we need to take full advantage of it.
And finally, as new companies join the community it is essential to educate them on how the project actually works and how they could contribute to it and, if needed, to change its course. Only then, we will have evolving and stable project that will thrive for years to come.
In the past few years, Linux kernel has undergone some significant changes, many of which have changed the way input/output works. It is therefore became a hot topic for DBAs and all those who work with databases.
In my talk, I will go through the recent changes - what was the reason for making them? what was the urgency for their implementation? and what does all that mean for databases and the way we work with them?
This topic has three major parts: Postgres IO, Linux IO and the interaction between the two. Due to the time constraints, I will not be able to dive deep into the details of basic Postgres IO, so I would expect some level of understanding of what checkpoint is, what does bgwriter etc. My talk will focus on the actual changes within Linux and how those changes affect Postgres.