Apache Storm 2.0.0 released over the past year, the new version is a significant improvement in performance integration, new features and the external system from its last update, here are some of the major features and improvements:
With a new architecture implemented in Java
In the previous version, Storm large part of the core functionality is implemented in Clojure in. Storm 2.0.0 has redesigned its core functions implemented in pure Java. The new Java-based implementation significantly improve the performance of the Storm's internal API easier to maintain and extend.
The new high-performance core
Storm 2.0.0 introduces a new core, it has a more streamlined threading model, ultra-fast and lightweight messaging subsystem backpressure model. It is designed to break through the boundaries of throughput, latency and power consumption, while maintaining backward compatibility. Storm 2.0 be the first to break the barrier of 1 microsecond latency streaming engine.
New Streams API
The new version introduces a new type of API, in order to more easily flow calculation expression. It is built on top of Storm's core spouts and bolt API, and automatic integration of more operations to optimize the pipeline.
Window Enhancements
Storm 2.0.0 The status window Windowing API can save / restore the state back-end configuration to support larger contiguous window. You can now access the window border via the API.
Integrated change Kafka
- Delete Storm-Kafka
- Migrating to KafkaConsumer.assign API
The end of the life cycle version 1.0.x.
Migrating to Java 8
other
- Enhanced SQL
- Improved Metrics
- New security features, such as nimbus group management, delegated token and optional analog
- Recombinant module dependency resolution and improved
- Improved API
- Support Lambda
- Enhanced resource-aware scheduler
- The new cluster management command for debugging state
Download: https://storm.apache.org/downloads.html
Apache Storm is a free open source distributed real-time computing systems. Simplifying the reliable stream data like Hadoop as real-time batch processing. Storm is very simple and can be used with any programming language. Apache Storm using Clojure development.
Storm has many application scenarios, including real-time data analysis, online learning, continuous computing, distributed RPC, ETL and so on. Storm is very fast, a test to achieve one million per second on a single set of processing nodes.
