Highlights FullContact: Improving the Graph: Transition to ScyllaDB FullContact set an ambitious goal of 10,000 QPS. Initially, they moved their database from HBase to Cassandra. Cluster consisted of 3 instances of c5.2xlarge EC2 + 2 TB of gp2 EBS storage. With the growing amount of records in the database, response time crept from 100 ms to 300 ms. It turned out that the default Size Tiered Compaction Strategy is optimized for inserts which lead to a single file for SSTable.

Highlights Slack: Migrating Millions of Concurrent Websockets to Envoy Slack makes an extensive use of websocket technology for their messaging service. Historically, they used HAProxy as a load balance, however, they faced an issue with dynamic updates of a list of endpoints. They could also change config and restart a load balancer which is tricky as it has to maintain existing websocket connections. They’ve decided to switch to Envoy proxy as it allows dynamic change of the configuration.