[ https://issues.apache.org/jira/browse/TINKERPOP-1492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15562827#comment-15562827 ]
ASF GitHub Bot commented on TINKERPOP-1492: ------------------------------------------- GitHub user okram opened a pull request: https://github.com/apache/tinkerpop/pull/453 TINKERPOP-1492: RemoteStrategy or the RemoteConnection should append a lazy barrier(). https://issues.apache.org/jira/browse/TINKERPOP-1492 I think that `LazyBarrierStrategy` should add a `barrier()` to the end of a traversal if it believes it will reduce the number of traversers. This is particularly important for `withRemote()` as you can greatly reduce the network traffic. I don't have a good way of knowing if the traversal is being executed remote or not so this will lazy `barrier()` will be added regardless of remote'ing. Please review and tell me if you disagree with where I am placing this optimization. VOTE +1. You can merge this pull request into a Git repository by running: $ git pull https://github.com/apache/tinkerpop TINKERPOP-1492 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/tinkerpop/pull/453.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #453 ---- commit 7c026acb905b68bddf46f0f6bb8501471e347d67 Author: Marko A. Rodriguez <okramma...@gmail.com> Date: 2016-10-10T16:50:54Z if the probability of bulking is high at the end step, a barrier(SIZE) is inserted. ---- > RemoteStrategy or the RemoteConnection should append a lazy barrier(). > ---------------------------------------------------------------------- > > Key: TINKERPOP-1492 > URL: https://issues.apache.org/jira/browse/TINKERPOP-1492 > Project: TinkerPop > Issue Type: Improvement > Components: driver, process, server > Affects Versions: 3.2.2 > Reporter: Marko A. Rodriguez > > I think that every remote traversal should have a {{barrier()}} appended to > the end. Why? It is cheaper to bulk and send a bulk over the wire than to > send multiple of the same traversers over the wire. > In essence, the same optimization we reach for with bulking for execution > should be used for bulking for serialization. > The problem/question I have is, where should this optimization be inserted? > 1. By GremlinServer (the remote connection) after received bytecode and > creating a traversal? > 2. By {{LazyBarrierStrategy}} that should just ALWAYS add a {{barrier()}} to > the end of every traversal? > 3. By {{RemoteStrategy}} by injecting {{barier()}}-bytecode right before > sending the bytecode over the wire? > I'm leaning on (2) given that its the natural way in which we currently do > things. -- This message was sent by Atlassian JIRA (v6.3.4#6332)