This is an automated email from the ASF dual-hosted git repository. mergebot-role pushed a commit to branch mergebot in repository https://gitbox.apache.org/repos/asf/beam-site.git
commit 708bd587b45934e52a12a282b107cb14c4b45171 Author: melissa <meliss...@google.com> AuthorDate: Thu Sep 28 18:19:12 2017 -0700 More updates from review feedback --- src/documentation/execution-model.md | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) diff --git a/src/documentation/execution-model.md b/src/documentation/execution-model.md index e09fe56..049c7f3 100644 --- a/src/documentation/execution-model.md +++ b/src/documentation/execution-model.md @@ -22,14 +22,8 @@ limit the distribution of output to other machines. ### Serialization and communication -The runner may serialize elements between machines for communication purposes -and for other reasons such as persistence. For example, a runner may serialize -and persist elements in these two situations. - -1. When used as part of a stateful `DoFn`, the runner may persist values to some - state mechanism. -1. When committing the results of processing, the runner may persist the outputs - as a checkpoint. +The runner might serialize elements between machines for communication purposes +and for other reasons such as persistence. A runner may decide to transfer elements between transforms in a variety of ways, such as: @@ -46,6 +40,13 @@ ways, such as: This may allow the runner to avoid serializing elements; instead, the runner can just pass the elements in memory. +Some situations where the runner may serialize and persist elements are: + +1. When used as part of a stateful `DoFn`, the runner may persist values to some + state mechanism. +1. When committing the results of processing, the runner may persist the outputs + as a checkpoint. + ### Bundling and persistence Beam pipelines often focus on "[embarassingly parallel](https://en.wikipedia.org/wiki/embarrassingly_parallel)" @@ -165,7 +166,7 @@ down and abandoned. If a failure to process an element in `ParDo2` causes `ParDo1` to re-execute, these two steps are said to be _co-failing_. -For this example, we will use the dependently parallel `ParDo`s from figure 4. +For this example, we will use the two `ParDo`s from figure 4. In figure 7, worker two successfully executes `ParDo1` on all elements in bundle B. However, the worker fails to process an element in bundle D, so `ParDo2` -- To stop receiving notification emails like this one, please contact "commits@beam.apache.org" <commits@beam.apache.org>.