Extreme Performance and Scalability with Near Caches

The aim of near caching is to provide a bridge between fast, in-memory, local caching and remote, massively scalable, Data Grids in such way that most recently or most frequently accessed data is quickly available while at the same time, clients still being able to transparently and seamlessly access the remote Data Grid when needed. Due to zero latency access provided for local data and the scalability offered by the possibility of going to a remote Data Grid, it's no wonder that this is one of the most demanded Infinispan patterns. In this talk, Galder will offer a detailed view of the pattern with a look at best practices for deploying it in your own environment. The talk will finish with a demo showing near caching in action!

Transcript

2. Extreme Performance and Scalability with Near Caches Galder Zamarreño Senior Software Engineer Red Hat, Inc 31st October 2011
3. Galder Zamarreño • R&D Engineer, Red Hat Inc. • Infinispan developer • 5+ years exp. with distributed data systems • Twitter: @galderz • Blog: zamarreno.com
4. Agenda • Introduction to Infinispan • Near caches • Clustered near caches • Near caches with JMS • Demo
5. Introducing
6. What is Infinispan? An in-memory, highly available, elastic, and open source (LGPL) data grid platform
7. Infinispan can be used as...
8. In-memory cache Boost performance caching data that’s frequently accessed, or hard to calculate...etc
9. Perfectly suited for it! Highly concurrent thanks to MVCC, has built-in eviction, pluggable persistence...etc
10. Embedded access •Client application and Infinispan live in same JVM •Zero Latency Access!
11. Infinispan meets CDI...
12. ... and JSR-107
13. Infinispan is not just a cache!
14. In-memory data grid It’s a Fast, Available, Distributed, Elastic data store, not just a cache!
15. Remote access •Via protocols •REST •Hot Rod
16. Why separate data tier? Manage/tune data tier independently, helps build stateless application tiers, and is easily scalable!
17. Data tier latency Network increases data grid access latency
19. Data tier
    + zero latency?
20. Near caches
21. Tiers in more detail
22. Near cache miss
23. Near cache hit
24. How to keep near caches consistent?
25. Cluster near caches
26. Server notifications Clustering near caches not always possible, so server to client notifications needed
27. JMS to the rescue! Hot Rod protocol not there yet, but JMS can easily fill the gap!
28. Near caches with JMS
29. Demo
30. Demo architecture
31. Summary Infinispan can be used as a fast, powerful local cache to boost performance
32. Summary Infinispan can also be used as a Fast, Available, Distributed, Elastic data grid
33. Summary Infinispan caches and data grids can be combined into performant, scalable near caches
34. Questions? http://infinispan.org http://blog.infinispan.org http://twitter.com/infinispan Rate this talk! http://speakerrate.com/galder