Skip to content
About
Main content
Model

Broadcast or shuffle a join

The shuffle's cost was relocated into memory, not removed.

Large side
the large input
Small side
the small input (copied to every node)
Parallelism
nodes

Recurring principle: Costs are conserved

How it behaves

One input on the horizontal axis, the others held fixed.

broadcastshuffle
Data table
Bytes over the network by parallelism (R = 400 GB, S = 500 MB)
Parallelism PseriesBytes moved
10broadcast4.66 GB
10.8broadcast5.03 GB
11.7broadcast5.44 GB
12.6broadcast5.88 GB
13.6broadcast6.35 GB
14.7broadcast6.87 GB
15.9broadcast7.42 GB
17.2broadcast8.02 GB
18.6broadcast8.67 GB
20.1broadcast9.37 GB
21.7broadcast10.1 GB
23.5broadcast10.9 GB
25.4broadcast11.8 GB
27.5broadcast12.8 GB
29.7broadcast13.8 GB
32.1broadcast14.9 GB
34.7broadcast16.1 GB
37.5broadcast17.4 GB
40.5broadcast18.9 GB
43.8broadcast20.4 GB
47.3broadcast22 GB
51.1broadcast23.8 GB
55.2broadcast25.7 GB
59.7broadcast27.8 GB
64.5broadcast30 GB
69.7broadcast32.5 GB
75.4broadcast35.1 GB
81.5broadcast37.9 GB
88broadcast41 GB
95.1broadcast44.3 GB
103broadcast47.9 GB
111broadcast51.8 GB
120broadcast55.9 GB
130broadcast60.4 GB
140broadcast65.3 GB
152broadcast70.6 GB
164broadcast76.3 GB
177broadcast82.5 GB
191broadcast89.1 GB
207broadcast96.3 GB
224broadcast104 GB
242broadcast113 GB
261broadcast122 GB
282broadcast131 GB
305broadcast142 GB
330broadcast154 GB
356broadcast166 GB
385broadcast179 GB
416broadcast194 GB
450broadcast210 GB
486broadcast226 GB
526broadcast245 GB
568broadcast264 GB
614broadcast286 GB
663broadcast309 GB
717broadcast334 GB
775broadcast361 GB
838broadcast390 GB
905broadcast422 GB
978broadcast456 GB
1.06Kbroadcast492 GB
1.14Kbroadcast532 GB
1.24Kbroadcast575 GB
1.33Kbroadcast622 GB
1.44Kbroadcast672 GB
1.56Kbroadcast726 GB
1.69Kbroadcast785 GB
1.82Kbroadcast848 GB
1.97Kbroadcast917 GB
2.13Kbroadcast991 GB
2.3Kbroadcast1.05 TB
2.49Kbroadcast1.13 TB
2.69Kbroadcast1.22 TB
2.9Kbroadcast1.32 TB
3.14Kbroadcast1.43 TB
3.39Kbroadcast1.54 TB
3.66Kbroadcast1.67 TB
3.96Kbroadcast1.8 TB
4.28Kbroadcast1.95 TB
4.63Kbroadcast2.1 TB
5Kbroadcast2.27 TB
10shuffle373 GB
10.8shuffle373 GB
11.7shuffle373 GB
12.6shuffle373 GB
13.6shuffle373 GB
14.7shuffle373 GB
15.9shuffle373 GB
17.2shuffle373 GB
18.6shuffle373 GB
20.1shuffle373 GB
21.7shuffle373 GB
23.5shuffle373 GB
25.4shuffle373 GB
27.5shuffle373 GB
29.7shuffle373 GB
32.1shuffle373 GB
34.7shuffle373 GB
37.5shuffle373 GB
40.5shuffle373 GB
43.8shuffle373 GB
47.3shuffle373 GB
51.1shuffle373 GB
55.2shuffle373 GB
59.7shuffle373 GB
64.5shuffle373 GB
69.7shuffle373 GB
75.4shuffle373 GB
81.5shuffle373 GB
88shuffle373 GB
95.1shuffle373 GB
103shuffle373 GB
111shuffle373 GB
120shuffle373 GB
130shuffle373 GB
140shuffle373 GB
152shuffle373 GB
164shuffle373 GB
177shuffle373 GB
191shuffle373 GB
207shuffle373 GB
224shuffle373 GB
242shuffle373 GB
261shuffle373 GB
282shuffle373 GB
305shuffle373 GB
330shuffle373 GB
356shuffle373 GB
385shuffle373 GB
416shuffle373 GB
450shuffle373 GB
486shuffle373 GB
526shuffle373 GB
568shuffle373 GB
614shuffle373 GB
663shuffle373 GB
717shuffle373 GB
775shuffle373 GB
838shuffle373 GB
905shuffle373 GB
978shuffle373 GB
1.06Kshuffle373 GB
1.14Kshuffle373 GB
1.24Kshuffle373 GB
1.33Kshuffle373 GB
1.44Kshuffle373 GB
1.56Kshuffle373 GB
1.69Kshuffle373 GB
1.82Kshuffle373 GB
1.97Kshuffle373 GB
2.13Kshuffle373 GB
2.3Kshuffle373 GB
2.49Kshuffle373 GB
2.69Kshuffle373 GB
2.9Kshuffle373 GB
3.14Kshuffle373 GB
3.39Kshuffle373 GB
3.66Kshuffle373 GB
3.96Kshuffle373 GB
4.28Kshuffle373 GB
4.63Kshuffle373 GB
5Kshuffle373 GB

Worked examples, recomputed

Every number the mechanisms quote for this relationship is recomputed from it, so none of them disagrees with it.

Where it appears

Each mechanism states it in its own names. The general symbol each name stands for is shown with it.

  • Query Processing

    Why can one question be answered by plans whose costs differ by a factor of ten thousand, and what does each operator force the engine to do?

    Read it in place →