Ultralow Latency with Java 11 and Terabytes of Data
summary
Claims retrieval of Java stream elements under 200 ns by mapping serialized rows into off-heap bytebuffers with off-heap indexes so garbage collection is unaffected. Demonstrates snapshots from a relational database into JVM RAM, incremental snapshot updates, and stream-aware short-circuiting so filters and aggregations run without creating heap objects. Reports microbenchmarks like ~6M ops/s (~175 ns) and ~14M with GraalVM (<100 ns). Not suited for very high-frequency transactional updates; deploys on large single hosts, Docker, or cloud VMs.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 1:12Ultralow Latency
- 2:31Why are application slow
- 4:15Example
- 8:09Context Switch
- 8:41Garbage Collect
- 9:23Unnecessary Object Creation
- 12:51CPU Consumption
- 13:18JVM Memory
- 15:45In Memory
- 16:10Scaling
- 17:45More than 12TB
- 19:15What if data grows
- 19:52Another simulation
- 20:26Speedman
- 22:53Small heap
- 23:13Java stream
- 28:19Parallel processing
- 28:57Handson demo
- 29:16Benchmark
- 29:49Initializer
- 30:23Demonstration
- 33:29Results
- 38:07Questions
- 40:33Memory Allocation
- 42:25Other Sources
- 45:02Object Mapping
- 45:26Heap Management
related talks
50mUltra fast Java In Memory Database Apps and Microservices with MicroStream
44mJava In-Memory Database-Apps with MicroStream
40mHigh-Performance Java-Cloud-Native Apps & Microservices
32mJava's Ignored Potential: Supercharge Database Apps with Java In-Memory Process
01h 12mHidden pearls for high-performance-persistence in Java
02h 57mPushing Java to the Limits: Processing a Billion Rows in under 2 Seconds
01h 01mJava compilers: Why Java is one of the first options for Low latency apps
54mEclipse MicroStream - Ultra-Fast Java Cloud-Native Persistence for Microservice