Binlog Server at Facebook
MySQL binary log management is one of the biggest challenges in MySQL clusters at the Facebook scale. Storing, indexing, retrieving and replaying binary logs is a complex task that we found ourselves re-implementing in our services over and over. To streamline our systems, we have designed a multi-purpose binlog management service called Binlog Server. Binlog Server can act as a replica and store binlogs, and it can also pretend to be a MySQL master and serve binlogs. In this talk, we'll show you a few of the use cases of Binlog Server at Facebook and cover its design details in depth. Highlights of the talk include: * Introduction to binary log storage at Facebook and the motivation behind Binlog Server * Design and implementation details * Examples of Binlog Server's use cases in MySQL automation tools at Facebook, such as handling cases when the master has purged binlogs slaves need, and using Binlog Server as a semi-sync replica. * Comparison with Max-scale's binlog server. Attendees will be able to fully understand the approach that Facebook has taken to manage MySQL binlog storage and how this new binlog server has affected our production efficiency. The target audience is anyone using MySQL replication extensively in their deployment.
Research Scientist, Facebook, Inc.
Teng is a Research Scientist in Facebook's database engineering team. His work is primarily focused on improving MySQL replication, by providing efficient MySQL binary log storage services used at the Facebook scale. Before joining Facebook, he graduated from the George Washington University with a Ph.D. in parallel and distributed computing.
Software Engineer, Facebook
Santosh is a software engineer at Facebook working on improving MySQL. He improved multiple replication features like Global Transaction Identifiers, Multi-threaded slave and Semi-sync so they work at Facebook scale efficiently. Before joining Facebook, he graduated from Indian Institute of Technology.