non deterministic result on LdbcSnbShortRead Generator due to replenish

[filipecosta90]

Hi there,
Considering that the inputs of short queries 1 to 7 are dependent upon the replenish of Complex queries output how can we ensure that two distinct runs will issue the same queries?
That said, If I replenish personIds and messageIds for the short Ids buffers from complex querie 2 it will have a different input if I replenish from complex querie 7.
Please denote https://github.com/ldbc/ldbc_snb_driver/blob/master/src/main/java/com/ldbc/driver/workloads/ldbc/snb/interactive/LdbcSnbShortReadGenerator.java#L531

With that in mind we can produce performance results with different values for the same queries that influenced by the complex queries.
Would it be better if we generated the input for short queries in the same manner we do it for Complex and BI ones?

[szarnyasg]

@filipecosta90 thanks for the suggestion, we'll discuss it tomorrow with the task force.

[szarnyasg]

Considering that the inputs of short queries 1 to 7 are dependent upon the replenish of Complex queries output how can we ensure that two distinct runs will issue the same queries?

Btw, given the same input parameters, the results of complex queries are deterministic. So why would two runs yield different results?

[filipecosta90]

Considering that the inputs of short queries 1 to 7 are dependent upon the replenish of Complex queries output how can we ensure that two distinct runs will issue the same queries?

Btw, given the same input parameters, the results of complex queries are deterministic. So why would two runs yield different results?

Hi there @szarnyasg , considering that for benchmark 1 we only enable Interactive Complex 2 and all short queries and for benchmark 2 we only enable Interactive Complex 7 and all short queries. Wont the input used for short queries vary?
We can keep determinism if we always use the same Complex query to replenish uppon but even on that case we're not able to ensure order ( only when using a single thread ).
What do you think?