The document summarizes updates and new features in Apache Storm 1.0 including:
- Pacemaker replaces Zookeeper for distributed coordination allowing larger clusters
- A distributed cache API allows sharing files between topologies
- High availability Nimbus allows Nimbus hosts to join/leave dynamically
- Native streaming windows support event time processing and out of order tuples
- State management provides automatic checkpointing for stateful bolts
- Automatic back pressure throttles spouts when buffers reach high watermarks
- The resource aware scheduler allocates tasks based on component resource requirements
- Usability improvements include dynamic log levels, tuple sampling for debugging, and distributed log searching
- New integrations support Cassandra, Sol
1 of 65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
More Related Content
The Future of Apache Storm
1. The Future of Apache Storm
Hadoop Summit 2016 - Dublin
P. Taylor Goetz, Hortonworks
@ptgoetz
4. Apache Storm 0.9.x
• First official Apache Release
• Storm becomes an Apache TLP
• 0mq to Netty for inter-worker communication
• Expanded Integration (Kafka, HDFS, HBase)
• Dependency conflict reduction (It was a start ;) )
13. Distributed Cache API
• Topology resources:
• Dictionaries, ML Models, Geolocation Data, etc.
• Typically packaged in topology jar
• Fine for small files
• Large files negatively impact topology startup time
• Immutable: Changes require repackaging and deployment
14. Distributed Cache API
• Allows sharing of files (BLOBs) among topologies
• Files can be updated from the command line
• Allows for files from several KB to several GB in size
• Files can change over the lifetime of the topology
• Allows for compression (e.g. zip, tar, gzip)
15. Distributed Cache API
• Two implementations: LocalFsBlobStore and HdfsBlobStore
• Local implementation supports Replication Factor (not needed for
HDFS-backed implementation)
• Both support ACLs
16. Distributed Cache API
Creating a blob:
storm blobstore create --file dict.txt --acl o::rwa
--repl-fctr 2 key1
Making it available to a topology:
storm jar topo.jar my.topo.Class test_topo -c
topology.blobstore.map=‘{"key1":{"localname":"dict.t
xt", "uncompress":"false"}}'
22. HA Nimbus
• Increase overall availability of Nimbus
• Nimbus hosts can join/leave at any time
• Leverages Distributed Cache API
• Topology JAR, Config, and Serialized Topology uploaded to
Distributed Cache
• Replication guarantees availability of all files
38. Automatic Back Pressure
• In previous Storm versions, the only way to throttle topologies was to
enable ACKing and set topology.spout.max.pending.
• If you don’t require at-least-once guarantees, this imposed a
significant performance penalty.**
** In Storm 1.0 this penalty is drastically reduced (more on this later)
39. Automatic Backpressure
• High/Low Watermarks (expressed as % of buffer size)
• Back Pressure thread monitors buffers
• If High Watermark reached, slow down Spouts
• If Low Watermark reached, stop throttling
• All Spouts Supported
41. Resource Aware Scheduler
• Specify the resource requirements (Memory/CPU) for individual
topology components (Spouts/Bolts)
• Memory: On-Heap / Off-Heap (if off-heap is used)
• CPU: Point system based on number of cores
• Resource requirements are per component instance (parallelism
matters)
42. Resource Aware Scheduler
• CPU and Memory availability described in storm.yaml on each
supervisor node. E.g.:
supervisor.memory.capacity.mb: 3072.0
supervisor.cpu.capacity: 400.0
• Convention for CPU capacity is to use 100 for each CPU core
43. Resource Aware Scheduler
Setting component resource requirements:
SpoutDeclarer spout = builder.setSpout("sp1", new TestSpout(), 10);
//set cpu requirement
spout.setCPULoad(20);
//set onheap and offheap memory requirement
spout.setMemoryLoad(64, 16);
BoltDeclarer bolt1 = builder.setBolt("b1", new MyBolt(), 3).shuffleGrouping("sp1");
//sets cpu requirement. Not neccessary to set both CPU and memory.
//For requirements not set, a default value will be used
bolt1.setCPULoad(15);
BoltDeclarer bolt2 = builder.setBolt("b2", new MyBolt(), 2).shuffleGrouping("b1");
bolt2.setMemoryLoad(100);
46. Dynamic Log Levels
• Set log level setting for a running topology
• Via Storm UI and CLI
• Optional timeout after which changes will be reverted
• Logs searchable from Storm UI/Logviewer
49. Tuple Sampling
• No more debug bolts or Trident functions!
• In Storm UI: Select a Topology or component and click “Debug”
• Specify a sampling percentage (% of tuples to be sampled)
• Click on the “Events” link to view the sample log.
50. Distributed Log Search
• Search across all log files for a specific topology
• Search in archived (ZIP) logs
• Results include matches from all Supervisor nodes
51. Dynamic Worker Profiling
• Request worker profile data from Storm UI:
• Heap Dumps
• JStack Output
• JProfile Recordings
• Download generated files for off-line analysis
• Restart workers from UI
52. Supervisor Health Checks
• Identify Supervisor nodes that are in a bad state
• Automatically decommission bad nodes
• Simple shell script
• You define what constitutes “Unhealthy”