MongoDB on AWS With Minimal Suffering
What should you know about running MongoDB on AWS? We'll cover a wide range of tips and tricks for squeezing the most performance out of your replica sets, as well as achieving maximum reliability and cost-effectiveness. For example: * What is the optimal replica set configuration for your workload? * When might you want to use arbiters? * What instance types should you use with MongoDB? * When should you use EBS vs PIOPS vs ephemeral storage? * How many PIOPS should you provision? What does capacity planning look like? * What are the best ways to perform backups? * Tips and tricks for improving performance (and some traps for the unwary) * Good and bad strategies for managing replica sets with code (Chef, Ansible, CloudFormation etc) * What are the most important things to monitor? * What are some AWS-related outage events you may encounter, and how can you gracefully recover from them? Some of these topics have very different answers depending on which storage engine you are running. We'll talk a little bit about how things like capacity planning, backup strategies, instance and storage type selection etc will be different when you are running WiredTiger or RocksDB under the hood instead of the classic old MMAPv1 storage engine. This is a fairly entry level talk, but there is a lot of useful stuff for people who are newer to Mongo or AWS. It may also be interesting to experts who are looking at switching to one of the newer storage engines, and are curious how it will impact their infrastructure.
Mike currently works at Truss where he is building systems and infrastructure for the common good. He comes from Parse, a mobile backend as a service, where he worked with a small team to manage one of the largest and most complex MongoDB footprints in the world. While at Parse, Mike worked on getting the first production instances of MongoDB with RocksDB as the underlying storage engine. This doubled Parse’s write capacity and dropped storage utilization by 10x. Mike was also part of the “Tech Surge” for healthcare.gov back when the initial rollout crumbled because there was no accountability, monitoring or communication. Mike graduated from University of California Santa Barbara in Computer Science.