Performance related Question

[tugan]

Hi,

we have single vertex and single edge very simple model and use dynamodb backend. increased table read capacity.

when we try yo get around 3000 record from server it takes 20-25 second. it that normal? should we do any additional configuration changes?

regards
Burak Tugan

[amcp]

The DynamoDB Storage Backend for Titan self-throttles to avoid getting throttled on the service side. Getting 3000 small (<1KiB) items in 20 seconds suggests that your read capacity is set to no more than 150 RPS, assuming that you use the single-item data model, no vertex partitioning and assuming that you read each record using random access (as opposed to a random scan). After you share details about your access pattern, data model and partitioning settings, and the throughput configuration of your graph, I could help you some more.

[amcp]

Also, you should definitely turn on query.fast-property=true, to pull all the properties of a vertex when it gets loaded (instead of lazy-loading them whenever they are referenced)

[amcp]

If you have any more questions about how to improve the query and traversal performance of Titan on DynamoDB, please re-open this issue.

[bendavidwhite]

Hi @tugan,

did the suggestions above improve your retrieval times?

I'm having similar performance issues but can't quite get my head around how resolve them (see #235 )