DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario #132396

tteofili · 2025-08-04T12:24:24Z

As discussed here, we might favor / penalize exploration of certain segments based on query vs segment affinity.
This does that by leveraging information about segment density (vectors per cluster), query to global centroid similarity.
Segments with higher affinity get increased visited_ratio, whereas segments with lower affinity see their visited_ratio decreased, optionally segments with very small affinity might not get explored.

…segment_affinity

…csearch into diskbbq_segment_affinity

…segment_affinity

…csearch into diskbbq_segment_affinity

…segment_affinity

…csearch into diskbbq_segment_affinity

…, minor improvements

…segment_affinity

…affinity as visited ratio modifier

benwtrent · 2025-08-25T18:23:45Z

I benchmarked this change, just using regular indexing & merge policy. I wonder if there is a bug? For very low percentage visits (1% or less), its worse.

Then for higher visit percentages, its only marginally better.

I also benchmarked with a 10% selectivity filter, and there is zero difference between the two implementations.

index_name                      index_type  num_docs  index_time(ms)  force_merge_time(ms)  num_segments
------------------------------  ----------  --------  --------------  --------------------  ------------
cohere-wikipedia-docs-768d.vec         ivf   8000000               0                     0            24
corpus-ann-gist-1M.fvec                ivf   1000000               0                     0            21
wiki1024en.train                       ivf   1000000               0                     0            18

BASELINE

index_name                      index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall    visited  filter_selectivity
------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ---------  ------------------
cohere-wikipedia-docs-768d.vec         ivf                 0.50         4.24              0.00           0.00  235.85    0.91   87476.40                1.00
cohere-wikipedia-docs-768d.vec         ivf                 1.00         6.92              0.00           0.00  144.51    0.93  166970.99                1.00
cohere-wikipedia-docs-768d.vec         ivf                 2.00        10.91              0.00           0.00   91.66    0.94  327095.79                1.00
corpus-ann-gist-1M.fvec                ivf                 0.50         1.23              0.00           0.00  813.01    0.88   15789.88                1.00
corpus-ann-gist-1M.fvec                ivf                 1.00         1.69              0.00           0.00  591.72    0.90   25413.18                1.00
corpus-ann-gist-1M.fvec                ivf                 2.00         2.39              0.00           0.00  418.41    0.90   45134.49                1.00
corpus-ann-gist-1M.fvec                ivf                 3.00         3.16              0.00           0.00  316.46    0.90   64945.82                1.00
corpus-ann-gist-1M.fvec                ivf                 4.00         4.01              0.00           0.00  249.38    0.90   85914.20                1.00
corpus-ann-gist-1M.fvec                ivf                 5.00         4.54              0.00           0.00  220.26    0.90  105406.93                1.00
wiki1024en.train                       ivf                 0.50         1.02              0.00           0.00  980.39    0.79   17694.47                1.00
wiki1024en.train                       ivf                 1.00         1.26              0.00           0.00  793.65    0.84   26217.51                1.00
wiki1024en.train                       ivf                 2.00         1.90              0.00           0.00  526.32    0.90   47006.85                1.00
wiki1024en.train                       ivf                 3.00         2.43              0.00           0.00  411.52    0.93   66709.42                1.00
wiki1024en.train                       ivf                 4.00         3.06              0.00           0.00  326.80    0.94   86651.58                1.00
wiki1024en.train                       ivf                 5.00         3.64              0.00           0.00  274.73    0.95  106544.51                1.00

CANDIDATE:

index_name                      index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall    visited  filter_selectivity
------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ---------  ------------------
cohere-wikipedia-docs-768d.vec         ivf                 0.50         6.68              0.00           0.00  149.70    0.92  129804.80                1.00
cohere-wikipedia-docs-768d.vec         ivf                 1.00         8.48              0.00           0.00  117.92    0.93  166970.99                1.00
cohere-wikipedia-docs-768d.vec         ivf                 2.00        10.36              0.00           0.00   96.53    0.94  241959.52                1.00
corpus-ann-gist-1M.fvec                ivf                 0.50         1.89              0.00           0.00  529.10    0.90   23801.30                1.00
corpus-ann-gist-1M.fvec                ivf                 1.00         2.05              0.00           0.00  487.80    0.90   25413.18                1.00
corpus-ann-gist-1M.fvec                ivf                 2.00         2.06              0.00           0.00  485.44    0.90   28470.45                1.00
corpus-ann-gist-1M.fvec                ivf                 3.00         2.88              0.00           0.00  347.22    0.90   40225.97                1.00
corpus-ann-gist-1M.fvec                ivf                 4.00         3.13              0.00           0.00  319.49    0.90   52354.58                1.00
corpus-ann-gist-1M.fvec                ivf                 5.00         3.66              0.00           0.00  273.22    0.90   63736.07                1.00
wiki1024en.train                       ivf                 0.50         1.20              0.00           0.00  833.33    0.81   20492.74                1.00
wiki1024en.train                       ivf                 1.00         1.53              0.00           0.00  653.59    0.84   26217.51                1.00
wiki1024en.train                       ivf                 2.00         1.83              0.00           0.00  546.45    0.88   38854.70                1.00
wiki1024en.train                       ivf                 3.00         2.40              0.00           0.00  416.67    0.91   55507.61                1.00
wiki1024en.train                       ivf                 4.00         3.34              0.00           0.00  299.40    0.93   71875.22                1.00
wiki1024en.train                       ivf                 5.00         3.45              0.00           0.00  289.86    0.94   87559.65                1.00

john-wagster · 2025-08-25T19:37:10Z

I benchmarked this change, just using regular indexing & merge policy. I wonder if there is a bug? For very low percentage visits (1% or less), its worse.

This makes sense. Where the PR is at currently we cap you at 1% as the lowest threshold. So likely what happens at 1% right now is we explore all segments at a 1% ratio at least but may explore more. So for configured 1% we're exploring too much. I can try to deal with this by saying we won't do this logic at all if we have a ratio that's below say 5%. So at 1% it would revert to the behavior on main.

        // for low affinity scores, decrease visited ratio
        if (affinityScore <= affinityThreshold) {
            return Math.max(visitRatio * 0.5f, 0.01f);
        }

Then for higher visit percentages, its only marginally better.

that surprises me given what @tteofili was showing me. Let me see if I can replicate and dig into those numbers.

I also benchmarked with a 10% selectivity filter, and there is zero difference between the two implementations.

I'll try to run with this as well.

…tio, lower all thresholds to favor smaller visit ratios

john-wagster · 2025-08-25T21:30:04Z

I cleaned up the magic numbers a bit. I'm seeing decent improvements with dbpedia. I'll run some additional datasets here later tonight.

I set the thresholds to be below half a percent where we cut off and lowered the min affinity from 1% to 0.1%. So you should see an actual impact at those levels you were testing @benwtrent.

dbpedia 4m (21-24 segments produced)

# baseline
index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 0.50         1.83              0.00           0.00  546.45    0.62    53792.41                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         3.04              0.00           0.00  329.22    0.70   100197.52                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 2.00         3.88              0.00           0.00  257.40    0.78   193119.21                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 3.00         4.86              0.00           0.00  205.87    0.81   285863.72                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 4.00         6.38              0.00           0.00  156.74    0.84   378586.81                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         7.26              0.00           0.00  137.65    0.86   471071.43                1.00

# candidate
index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 0.50         1.61              0.00           0.00  622.08    0.61    44239.54                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         2.87              0.00           0.00  348.74    0.68    78858.63                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 2.00         3.35              0.00           0.00  298.73    0.75   149642.79                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 3.00         4.60              0.00           0.00  217.27    0.79   219955.92                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 4.00         5.34              0.00           0.00  187.35    0.81   290437.33                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         5.64              0.00           0.00  177.38    0.84   360858.71                1.00

john-wagster · 2025-08-26T05:42:44Z

I ran dbpedia again with the same ingest this time. ~~And got good numbers still~~ I got confusing numbers. Low visit percentages have poor QPS but visit less, which doesn't make any sense. I'll look into that further:

# baseline
index_name                             index_type  num_docs  index_time(ms)  force_merge_time(ms)  num_segments
-------------------------------------  ----------  --------  --------------  --------------------  ------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf  10000000               0                     0            24

index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 0.50         1.79              0.00           0.00  558.66    0.63    54701.69                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         2.42              0.00           0.00  413.22    0.71   101428.57                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 2.00         3.58              0.00           0.00  279.72    0.78   194028.14                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 3.00         4.80              0.00           0.00  208.22    0.82   286612.36                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 4.00         5.89              0.00           0.00  169.78    0.84   379281.03                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         7.11              0.00           0.00  140.75    0.86   472046.09                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                10.00        12.31              0.00           0.00   81.27    0.91   935829.19                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                30.00        33.58              0.00           0.00   29.78    0.96  2790057.65                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                50.00        54.65              0.00           0.00   18.30    0.97  4644486.01                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                70.00        75.55              0.00           0.00   13.24    0.98  6498687.84                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf               100.00       107.76              0.00           0.00    9.28    0.98  9271902.00                1.00


# candidate
index_name                             index_type  num_docs  index_time(ms)  force_merge_time(ms)  num_segments
-------------------------------------  ----------  --------  --------------  --------------------  ------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf  10000000               0                     0            24

index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 0.50         2.14              0.00           0.00  466.74    0.61    44239.54                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         2.15              0.00           0.00  465.12    0.68    78858.63                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 2.00         3.04              0.00           0.00  328.68    0.75   149642.79                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 3.00         4.01              0.00           0.00  249.07    0.79   219955.92                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 4.00         4.86              0.00           0.00  205.97    0.81   290437.33                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         5.55              0.00           0.00  180.10    0.84   360858.71                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                10.00         9.81              0.00           0.00  101.88    0.89   713512.61                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                30.00        26.00              0.00           0.00   38.47    0.94  2123247.04                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                50.00        42.69              0.00           0.00   23.42    0.96  3533203.26                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                70.00        58.48              0.00           0.00   17.10    0.97  4943038.51                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf               100.00        83.27              0.00           0.00   12.01    0.98  7053592.12                1.00

However, when I started to run cohere I got similar results to what you originally reported @benwtrent. I'm seeing when filtering very little difference between candidate and baseline. And when not filtering I'm seeing the baseline sometimes doing better. I'm going to run it a few more times tomorrow. I see a lot of variation with the default merge policy across a few runs I did, which was a little surprising. And I'll see if there's anything that can be done about improvements here and post some additional numbers.

benwtrent · 2025-08-26T12:24:56Z

I see a lot of variation with the default merge policy across a few runs I did, which was a little surprising.

The concurrent merge scheduler may or may not kick off a merge due to thread availability and memory pressure. So, its not totally surprising that the number of segments jump between tier sizes (e.g. 5-10 in count)

…segment_affinity

…rease rather than increase

…csearch into diskbbq_segment_affinity

tteofili · 2025-09-05T14:30:30Z

I got to stabilize the results such that this can work both with the default (tiered) merge policy and with no merge policy, without breaking recall.
One problem in the earlier impl is that density was used to boost affinity whereas the number of vectors in a segment is a more important factor, especially when the size of segments has a higher variance (e.g., typical log-like distributions of tiered merge policy).
as a side note, to have comparable runs, the index has to be the same (hence I had to set reindex=false in the json config).

Cohere

Main

index_name       index_type  num_docs  index_time(ms)  force_merge_time(ms)  num_segments
---------------  ----------  --------  --------------  --------------------  ------------  
wiki1024en.docs         ivf   1000000               0                     0            18

index_name       index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count      QPS  recall    visited  filter_selectivity
---------------  ----------  -------------------  -----------  ----------------  -------------  -------  ------  ---------  ------------------  
wiki1024en.docs         ivf                 1.00         0.90              2.97           3.30  1111.11    0.91   30116.00                1.00
wiki1024en.docs         ivf                 5.00         0.85              3.77           4.44  1176.47    0.97   59804.29                1.00
wiki1024en.docs         ivf                10.00         1.21              5.23           4.32   826.45    0.99  108649.49                1.00
wiki1024en.docs         ivf                30.00         2.23              9.35           4.19   448.43    1.00  307106.92                1.00
wiki1024en.docs         ivf                50.00         3.42             13.91           4.07   292.40    1.00  506324.92                1.00
wiki1024en.docs         ivf                70.00         4.93             19.10           3.87   202.84    1.00  704843.20                1.00
wiki1024en.docs         ivf               100.00         6.44             25.29           3.93   155.28    1.00  998499.00                1.00

Candidate

index_name       index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count      QPS  recall    visited  filter_selectivity
---------------  ----------  -------------------  -----------  ----------------  -------------  -------  ------  ---------  ------------------  
wiki1024en.docs         ivf                 1.00         0.77              2.82           3.66  1298.70    0.91   30118.81                1.00
wiki1024en.docs         ivf                 5.00         0.82              3.44           4.20  1219.51    0.97   55186.97                1.00
wiki1024en.docs         ivf                10.00         1.12              4.11           3.67   892.86    0.98   98530.78                1.00
wiki1024en.docs         ivf                30.00         2.21              7.05           3.19   452.49    1.00  274222.32                1.00
wiki1024en.docs         ivf                50.00         3.38             10.53           3.12   295.86    1.00  451666.17                1.00
wiki1024en.docs         ivf                70.00         4.66             13.67           2.93   214.59    1.00  628783.98                1.00
wiki1024en.docs         ivf               100.00         6.54             18.75           2.87   152.91    1.00  891313.12                1.00

DBPedia

Main

index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         2.01              0.00           0.00  498.13    0.67    99594.35                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         6.45              0.00           0.00  155.10    0.84   470410.90                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                10.00        11.79              0.00           0.00   84.84    0.89   933995.92                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                30.00        33.64              0.00           0.00   29.73    0.93  2788274.12                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                50.00        53.81              0.00           0.00   18.58    0.94  4642552.50                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                70.00        74.87              0.00           0.00   13.36    0.95  6496735.59                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf               100.00       108.33              0.00           0.00    9.23    0.95  9271553.00                1.00

Candidate

index_name                             index_type  visit_percentage(%)  latency(ms)  net_cpu_time(ms)  avg_cpu_count     QPS  recall     visited  filter_selectivity
-------------------------------------  ----------  -------------------  -----------  ----------------  -------------  ------  ------  ----------  ------------------  
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 1.00         1.76              0.00           0.00  568.18    0.66    95574.59                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                 5.00         5.92              0.00           0.00  168.99    0.84   450293.62                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                10.00        11.27              0.00           0.00   88.69    0.89   893695.94                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                30.00        31.63              0.00           0.00   31.62    0.93  2667241.42                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                50.00        52.16              0.00           0.00   19.17    0.94  4440804.43                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf                70.00        71.50              0.00           0.00   13.99    0.95  6214189.25                1.00
corpus-dbpedia-entity-E5-small-0.fvec         ivf               100.00        99.96              0.00           0.00   10.00    0.95  8722690.54                1.00

the final result is that the reduction in visited_nodes is less evident, so while the idea seemed interesting, it's currently probably not worth the complexity.

tteofili added 11 commits July 31, 2025 16:04

DiskBBQ - adjust query greediness based on per segment affinity

6c508e3

wip fixes

6281bcf

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

62e4d0c

…segment_affinity

wip fixes

ab11958

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

980940e

…segment_affinity

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

3f4e5fd

…segment_affinity

Include top 2 parent scores into affinity, with larger segments

df0210d

minor fix

8cbad8f

include (parent) centroids scores for affinity, with larger segments

b097c20

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

9499be1

…segment_affinity

more sensible and generic threshold definition

0635453

elasticsearchmachine added the v9.2.0 label Aug 4, 2025

elasticsearchmachine and others added 18 commits August 4, 2025 12:36

[CI] Auto commit changes from spotless

d52dc40

Merge branch 'diskbbq_segment_affinity' of github.com:tteofili/elasti…

c658856

…csearch into diskbbq_segment_affinity

spotless

34cf456

minor tweaks

0dc22eb

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

028482e

…segment_affinity

minor tweaks

a6702eb

[CI] Auto commit changes from spotless

5288b8e

set visited vector max budget ratio

0a74a0c

Merge branch 'diskbbq_segment_affinity' of github.com:tteofili/elasti…

7600d50

…csearch into diskbbq_segment_affinity

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

c3cec18

…segment_affinity

minor

4f98473

minor tweaks

72d6d24

minor

9ee4bfd

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

4f9982e

…segment_affinity

minor tweaks, add knn tester param

aeee542

[CI] Auto commit changes from spotless

5facc1b

minor tweaks

feedbdf

Merge branch 'diskbbq_segment_affinity' of github.com:tteofili/elasti…

58a397b

…csearch into diskbbq_segment_affinity

john-wagster and others added 6 commits August 14, 2025 22:56

calculating affinities accounting for some of the wide variance seen

ff4916d

calculating affinities accounting for some of the wide variance seen

f3afcd6

don't add task with 0 budget, adjust thresholds for skewed affinities…

22ce362

…, minor improvements

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

480fc0d

…segment_affinity

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

e8dbdef

…segment_affinity

comment

d631243

tteofili mentioned this pull request Aug 21, 2025

DiskBBQ to dynamically adapt nProbe when centroids scores diff is large-ish (multi segment case) #131911

Closed

tteofili changed the title ~~DiskBBQ - Adapt nProbe based on query - segment affinity in multi segment scenario~~ DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario Aug 21, 2025

tteofili and others added 6 commits August 22, 2025 15:29

remove explicit budget as it conflicts with visitedRatio, normalized …

d261f02

…affinity as visited ratio modifier

don't use affinity calculation for filters

e9d4d82

minor fix

3b291df

more sensible visited ratio adjustment

76c22b9

removed useless interface

7f2e954

Merge branch 'main' into diskbbq_segment_affinity

d887b88

pull out magic nums, clean up, set hard cap on when to apply visit ra…

e8e7d12

…tio, lower all thresholds to favor smaller visit ratios

Merge branch 'main' into diskbbq_segment_affinity

a67fb67

tteofili and others added 7 commits September 3, 2025 11:38

Merge branch 'main' of github.com:elastic/elasticsearch into diskbbq_…

daea67d

…segment_affinity

merge

571534c

[CI] Auto commit changes from spotless

8e13a3a

account for numVectors for segment size variance being high, more dec…

3d1c6b6

…rease rather than increase

work better with unbalanced segments, simplified

daa8f40

Merge branch 'diskbbq_segment_affinity' of github.com:tteofili/elasti…

59cae53

…csearch into diskbbq_segment_affinity

changelog

b9a9e23

tteofili closed this Sep 5, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario #132396

DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario #132396

Uh oh!

tteofili commented Aug 4, 2025 •

edited

Loading

Uh oh!

benwtrent commented Aug 25, 2025

Uh oh!

john-wagster commented Aug 25, 2025

Uh oh!

john-wagster commented Aug 25, 2025 •

edited

Loading

Uh oh!

john-wagster commented Aug 26, 2025 •

edited

Loading

Uh oh!

benwtrent commented Aug 26, 2025

Uh oh!

tteofili commented Sep 5, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario #132396

DiskBBQ - Adapt visited_ratio based on query - segment affinity in multi segment scenario #132396

Uh oh!

Conversation

tteofili commented Aug 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

benwtrent commented Aug 25, 2025

Uh oh!

john-wagster commented Aug 25, 2025

Uh oh!

john-wagster commented Aug 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

john-wagster commented Aug 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

benwtrent commented Aug 26, 2025

Uh oh!

tteofili commented Sep 5, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

tteofili commented Aug 4, 2025 •

edited

Loading

john-wagster commented Aug 25, 2025 •

edited

Loading

john-wagster commented Aug 26, 2025 •

edited

Loading