格雷姆林遍历膨胀（OLTP）

Question

While i understand the principle of Bulking, i am curious to know how does the machine make that happen from a programatic point of you. How to detect that multiple Traverser are on the same node, to decide to merge and bulk them. Is there any description somewhere of the algorithm used for detecting that a specific situation requires a Bulk.

Also, what are the explicits Step that would trigger a bulk. Is there any instructions that can be added to the traversal that will make the bulk happen explicitly ?

Answer 1

Bulking typically occurs in relation to a barrier() step and is discussed a bit in TinkerPop's reference documentation here . Most folks don't use barrier() directly in their Gremlin writing (though you can) as various Traversal Strategies built into TinkerPop (like LazyBarrierStrategy as an example) try to automatically inject them into the traversal in places where they might do some good.

You can see where/when barrier() steps are being injected by doing an explain() on your traversal:

gremlin> g.V().out().both().explain()
==>Traversal Explanation
=======================================================================================================================================
Original Traversal                 [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]

ConnectiveStrategy           [D]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
IncidentToAdjacentStrategy   [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
MatchPredicateStrategy       [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
RepeatUnrollStrategy         [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
PathRetractionStrategy       [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
FilterRankingStrategy        [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
InlineFilterStrategy         [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
AdjacentToIncidentStrategy   [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
CountStrategy                [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), VertexStep(BOTH,vertex)]
LazyBarrierStrategy          [O]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]
TinkerGraphCountStrategy     [P]   [GraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]
TinkerGraphStepStrategy      [P]   [TinkerGraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]
ProfileStrategy              [F]   [TinkerGraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]
StandardVerificationStrategy [V]   [TinkerGraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]

Final Traversal                    [TinkerGraphStep(vertex,[]), VertexStep(OUT,vertex), NoOpBarrierStep(2500), VertexStep(BOTH,vertex)]

Answer 2

在这里，在上面的@stephen评论之后，我只想分享一个博客的链接，其中一个博客深入解释了著名的LazyBarrierStrategy