Oracle Coherence
- 1. Oracle CoherenceBy Mustafa Ahmed1
- 2. AgendaThe ProblemSolutionData GridsReplicated TopologyPartitioned TopologyNear TopologyEventsQueryRead Through CachingWrite Through CachingWrite Behind CachingCoherence Code ExamplesConclusionPerformance
- 3. Availability
- 5. DataExtreme increase in Access Volume & Complexity of DataDriving Data DemandVirtualizationAbility to move applications around several machinesService Oriented Architecture (SOA)Integrated services that can be used in Multiple business domainsRelying on other servicesThe Problem 3
- 6. Provide Reliable, Scalable, Universal Data Access and Management. PerformanceSolves Latency and Bandwidth ProblemsAvailabilityHaving the data available at all timesScalability Handle growing demand of Data Efficiently4Solution – Oracle Coherence
- 7. Manages Information in a grid environmentLots of servers working together
- 8. Servers do not run independently Server manages stateEven server failure occurs.Adding more serversConcept of scale outIt will manage more data and can handle more transactions per second. Data as a ServiceMiddle Tier
- 9. In App ServerData Integration is in Data ServiceIntegration can occur in Domain Model5Data Grids
- 10. Combines Data Management with Data ProcessingPush processing where data is being managed
- 11. Read or Write data across any number of serversSingle System ImageNo need to show server infrastructure
- 12. Pretend all the information is Local
- 13. Logical view of all data in all the servers6Data Grids
- 14. There are two things you can move in a Distributed EnvironmentStateDistribution of a state is referred to as replicationBehaviorMoving messagesData Grids combine these two conceptsYou can either move data or the processing where data is sitting
- 15. Push all the processes to the Information7Data Grids
- 16. Locality of DataMost applications spend most of the time waiting for data
- 17. If the data is partitioned with non overlapping regions the behavior can be moved to the server that owns the data to process
- 18. Results In lower latency8Data Grids
- 19. Technology introduced In 2001Replicate information among all serversData is replicated to all members in Data GridProblemsScalability ProblemCapacity of Information Stays the same9Replicated Topology
- 20. 10Replicated TopologyExpensiveUpdate Each Server every time
- 21. Conceptually its expensive Each Information is spread out across the servers (Peer to Peer)Load BalancerKeeps track of the load
- 22. Move from one server to another
- 23. Sends to the server which owns the dataExactly one server owns the informationHas a sync back up for it11Partitioned Topology
- 24. 12Partitioned TopologyFailure OccursThe operation still finishes correctly
- 25. Increase servers from 2 to 2000 servers it increases scalability
- 26. All servers are disposable at any period of timeL2 Cache vs. L1 CachePartitioned Topology as L2 Cache
- 27. Near Topology as L1 CacheStores it LocallyIf asks again then gets it locallyDemand base replicated cachingZero Latency access to recently used data13Near Topology
- 28. 14Near Topology
- 29. 15EventsAll the dataset provide events regardless of TopologyEvents are distributed efficiently to the interested listeners
- 30. Parallel QueryQuery performed parallel across the data grid using indexing
- 31. All doing the local portion of the Query16Query
- 32. Continuous QueryCombines a Query with Events to provide a local materialized view
- 33. Result is up to date in real time
- 34. Like in near topology but always contains the desired data17Query
- 35. 18Read Through CachingFinds it in L1 or L2 CacheOtherwise sends a request to the databaseOnly sends one requestsCoalesces multiple reads to reduce the database load
- 36. Writes first to the database and then commits to the cacheNot a Two-Phase CommitKeeps the in-memory data and the database in sync.19Write Through Caching
- 37. First writes it to the cacheLater commits it to the databaseThis assures the latest version of the cacheBatches all the writes into one objectGeico uses itImproved performance90% reduction in database usage20Write Behind Caching
- 38. Joins an existing cluster or forms a new oneLeaves the current cluster21Coherence Code ExamplesCluster cluster = CacheFactory.ensureCluster();CacheFactory.shutdown();
- 39. 22Coherence Code ExamplesNamedCachenc = CacheFactory.getCache(“mine”);Object previous = nc.put(“key”, “hello world”);Object current = nc.get(“key”);int size = nc.size();boolean exists = nc.containsKey(“key”);
- 40. Observe changes in real time as the occur23Coherence Code ExamplesNamedCachenc = CacheFactory.getCache(“stocks”);nc.addMapListener(new MapListener() { public void onInsert(MapEventmapEvent) { } public void onUpdate(MapEventmapEvent) { } public void onDelete(MapEventmapEvent) { } });
- 41. PerformanceSolves Latency Problems And Preserve network bandwidthCache recently used dataAbility to execute tasks parallel across the data grid Moving the process where the data is24Conclusion - Performance
- 42. AvailabilityRemove all single point of failureAdded redundancy to improve availabilityAble to Queue updates if database is not availableIncrease availability from 11 days to 2.5 hours per year25Conclusion - Availability
- 43. ScalabilityScale Out functionalityDatabase ShardingCoherence eliminates Database ShardingDistributed cache Updates performed against the cache dataScaling both capacity and throughput Adding more nodes to the Coherence Cluster26Conclusion - Scalability
- 44. Any Questions 27