Edge Performance with In memory NoSQL
summary
Compares three real deployments and what breaks: memcache as a simple client-side cache for sessions; Redis used as the only database in a demo showing lists/hashes and that long commands block its single thread (Sentinel polls every ~40s so blocking commands trigger failovers); Aerospike demo shows failed query until a secondary index is created, warns SSD volumes must be zeroed, indexes live in RAM (6–7 hour rebuild on restart), synchronous replication dislikes cross-AZ and network-poor t2 instances.
AI-written summary · about
chapters (18)
- 0:00Start
- 1:04About me
- 3:53Scalability at Plug And Travel?
- 7:34Autoscaling solutions
- 9:25Questions to have answers for
- 9:53First Question
- 12:42Meet the main characters
- 13:14Memcached-the veteran
- 15:09Redis-the mature
- 16:12Going with Redis
- 18:40Redis-Data structures
- 23:56Redis-pay attention to
- 26:54Aerospike-the new guy
- 28:26Going with Aerospike
- 30:49Aerospike-Storage models
- 37:19Aerospike-pay attention to
- 42:09Lessons learned @ PlugAndTravel
- 43:42Small review
related talks
38mAlternatives to Redis while not compromising on its speed
28mAerospike's Journey Through Innovations, Sustainability, and Roadmap
40mAerospike Demo Showcase
22mThe Art of the Possible
26mLeveraging Moore's Law to Optimize Database Performance
53mLeveraging Caching For Cloud Native Applications
47mCaching beyond simple put and gets
15mAWS On Air Voice of the Customer ft. Aerospike, Inc.