Multi-master replication is the holy grail of distributed applications. However, as applications scale over distances and data volumes, the hard problem is not replication. It's your application. This talk will use a simple but compelling example of airline bookings to show why replication is just plumbing and why application design is what makes multi-master really work.
Cruel things that you can inflict to your replicated cluster ... and get away with it!
You have heard about high availability systems. Maybe you have got one yourself. But unless you have had a serious failure, you probably don't know what disgraces your dbms can survive.
This talk shows, with a live demo, the kind of tortures you can inflict to a replicated cluster, which always survives despite your cruelty.
Pull the carpet under the master while an application is writing by promoting a slave without notice, pull the plug to the master database server and watch the system recover, kill connections and unmount disks. This is high availability in practice. Come and watch, unless you have a weak heart!
Diagnosis and Possible Fixes: a Multiple Choice Primer for Beginning DBAs
A tongue-in-cheek presentation of issues found in MySQL installations, in Q & A format with multiple choice questions and a little audience participation.
What to do with a slow server?
a. Check out your slow query log
b. add memory
c. throw it away
d. ask IT
Swap memory always being used up?
a. increase swap size
b. check what is running on your server
c. check your config files
d. reduce swap to 0 and see what happens
These and 18 other choice questions will be addressed for the inquisitive DBA.
If you have always wondered about some of the things that happen at your shop with MySQL and were afraid to ask, you may be right not to go ahead and ask. But at least you can attend this lightning talk and relate a little bit to the issues presented here.
Many books have been written about how to design scalable systems. People build their careers on this topic. What can I possibly say in five minutes that no one else has said? I have two big points to make, stunningly obvious once you know them, but completely under most people's radar. It may not shatter everything you think you know about scalability, but it will probably frame it all in a new context.
MAKE.DATABASES.FAST - MySQL server installation automation
The power of MySQL replication encourages horizontal database scaling. Booking.com uses exclusively MySQL to power a world-leading hotel reservation business. We will describe our automation for installing, configuring and copying databases to make new database instances with minimal effort and no DBAs! We use Kickstart, Puppet and scripts to configure hardware, install OS and software, and copy databases with minimal effort to grow our MySQL installation as fast as we need to.
The real end-game for Big Data is to have transactional and analytic data on the same database. Just imagine the maximum value one could get out of designing analytics as part of operations.
Databases used to be all-serving 40 years ago with one type of indexing structure used ubiquitously. This all worked well with the simple drives and application of the time. As the decades wore on, applications, storage, and database needs evolved, and the basic database diverged into all sorts of speciality products (ie, column store, NoSQL, etc...) to react. However, it needn't be this way.
We'll walk you through the decades and show the path from single database to pluralism back to single database. All in 5 minutes!
You find a column named EntityNum in a table you manage, but what data belongs in this column? Not every detail of usage is clear from just SQL data type and constraints. What is the sensible range of values? Unit of measure? How is the column used by applications? Who in the world knows? We need a way to add comments to the database schema, just as we would write comments in application code to document how programmers should use it. But comments are useful only if they're correct and current, and if they're easy to read and to update. Schemadoc is an experimental tool to help in these goals.