Lambda architecture for real-time big data processing

Problem: After the
results of Batch View and Realtime View are merged, the real result should be merged and all realtime and batch view results of the corresponding
task should be merged at this point after the batch recalculation starts, and the results of the realtime view should be cleared. 0 (locked, new data cannot be calculated at this time to prevent dirty data results), then unlocked to start batch calculation, and realtime view also began to calculate. Thinking: Is it possible to use a result? realtime directly updates the result of the batch (using zookeeper as a global lock, both sides must obtain the lock and then update) http://www.2cto.com/kf/201505/402080.html http://m.blog.csdn .net/blog/GreatElite/25502203



Guess you like

Origin http://10.200.1.11:23101/article/api/json?id=326803453&siteId=291194637