[
https://issues.apache.org/jira/browse/SOLR-13132?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17152287#comment-17152287
]
Michael Gibney commented on SOLR-13132:
---------------------------------------
Got it, yes, thanks! Merged cleanly, and eliminated (or suppressed as
selectively as possible) a handful of warnings unique to this branch (just
pushed).
Regarding the informal benchmarks, I haven't run them in some time, but it's
difficult to compare apples to apples. Full domain skg without sweep over
reasonably high-cardinality fields (and without the selective caching
introduced by SOLR-13108) currently easily causes filterCache thrashing (really
bad); a trivial alternative would be to not consult the filterCache at all, but
that carries a different (if more predictable) type of performance hit.
So in a sense it feels misleading to run comparison performance benchmarks
because non-sweep can't "put it's best foot forward" and a lot of the latency
you'd be measuring would be dependent on system-level stuff (filterCache
configuration and usage patterns, etc.).
I'll try to pull something together that encapsulates these variables in an
illustrative way: comparing with filterCache sufficient to hold all terms,
filterCache disabled, filterCache thrashing (probably worst-case, and most
realistic), and for different cardinality fields and DocSet domains. But
ultimately I think the big win is the consistent performance and system-wide
benefits that accompany a clean separation from filterCache – and setting up
for a bigger win (which I'll mention and then duck!) if/when sweep collection
can be integrated with term facet count cache (SOLR-13807).
> Improve JSON "terms" facet performance when sorted by relatedness
> ------------------------------------------------------------------
>
> Key: SOLR-13132
> URL: https://issues.apache.org/jira/browse/SOLR-13132
> Project: Solr
> Issue Type: Improvement
> Components: Facet Module
> Affects Versions: 7.4, master (9.0)
> Reporter: Michael Gibney
> Priority: Major
> Attachments: SOLR-13132-with-cache-01.patch,
> SOLR-13132-with-cache.patch, SOLR-13132.patch, SOLR-13132_testSweep.patch
>
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> When sorting buckets by {{relatedness}}, JSON "terms" facet must calculate
> {{relatedness}} for every term.
> The current implementation uses a standard uninverted approach (either
> {{docValues}} or {{UnInvertedField}}) to get facet counts over the domain
> base docSet, and then uses that initial pass as a pre-filter for a
> second-pass, inverted approach of fetching docSets for each relevant term
> (i.e., {{count > minCount}}?) and calculating intersection size of those sets
> with the domain base docSet.
> Over high-cardinality fields, the overhead of per-term docSet creation and
> set intersection operations increases request latency to the point where
> relatedness sort may not be usable in practice (for my use case, even after
> applying the patch for SOLR-13108, for a field with ~220k unique terms per
> core, QTime for high-cardinality domain docSets were, e.g.: cardinality
> 1816684=9000ms, cardinality 5032902=18000ms).
> The attached patch brings the above example QTimes down to a manageable
> ~300ms and ~250ms respectively. The approach calculates uninverted facet
> counts over domain base, foreground, and background docSets in parallel in a
> single pass. This allows us to take advantage of the efficiencies built into
> the standard uninverted {{FacetFieldProcessorByArray[DV|UIF]}}), and avoids
> the per-term docSet creation and set intersection overhead.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
