Page History

...

• Side Input Caching using tokens. 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key BEAM-11097

• Side inputs are valid based on the window. Multiple elements and bundles processed in the same window can reuse the same side input data at the runner’s discretion. This is implemented with a key token. 
• This is critical for Streaming performance to reduce data interchange with the Runner half of a worker.
  
• Map Side Inputs / Random Access Side Inputs. 

  Jira
  server ASF JIRA serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key BEAM-3293

• KV PCollections as side inputs currently support an iterable approach, however the beam protocol does support a Key Lookup approach as well.
• This enables having significantly larger side inputs, as data can be looked up from the runner as needed, rather than possibly iterating over the entire PCollection for every element.
• Similarly, an advanced feature would be to support custom ViewFns for side inputs user side processing of side inputs, orchestrated and garbage collected by the framework, rather than requiring users to do so manually.   

  Jira
  server ASF JIRA columns key,summary,type,created,updated,due,assignee,reporter,priority,status,resolution serverId 5aa69414-a9e9-3523-82ec-879b028fb15b key BEAM-110993305

Windowing [Basic]

FlumeGo supports simple fixed interval windowing, and global windows. The validation mentioned above needs to be handled before sliding windows can be declared to be correctly supported.

...