15.10 - Join Plan With Star Join Optimization and Fact Table Subquery Join - Teradata Database

Teradata Database SQL Request and Transaction Processing

prodname
Teradata Database
vrm_release
15.10
category
Programming Reference
User Guide
featnum
B035-1142-151K

The following query is used for this example.

     SELECT …
     WHERE widgets.color=COLOR.code
     AND   widgets.size=SIZE.code
     AND   widgets.options IN (SELECT OPTIONS.code);

With join optimization, the following join plan is generated when the collection of join columns (color, size, and options) makes up a nonunique secondary index of the large table:

 

                Operation

             Joined Tables

     Total Processing Time (seconds)

Spool 4: Product Join

duplicated color, direct size

                            0.31

Spool 6: Product Join

local 4, duplicated options

                            4.46

Spool 1: Nested Join

duplicated 6, index widgets

                          22.73

The total estimated completion time is 27.40 seconds.

The estimated performance improvement factor is 245.

The EXPLAIN output for this optimized join plan is as follows:

Explanation
--------------------------------------------------------------------
1) First, we lock TEST.Options for read, we lock TEST.Color for read,
   we lock TEST.Size for read, and we lock TEST.Widgets for read.
2) Next, we execute the following steps in parallel.
   a) We do an all-AMPs RETRIEVE step from TEST.Options by way of
      an all-rows scan with no residual conditions into Spool 2, which
      is redistributed by hash code to all AMPs. Then we do a SORT to
      order Spool 2 by the sort key in spool field1 eliminating
      duplicate rows. The size of Spool 2 is estimated to be 10 rows.
      The estimated time for this step is 0.19 seconds.
   b) We do an all-AMPs RETRIEVE step from TEST.Color by way of an
      all-rows scan with no residual conditions into Spool 3, which is
      duplicated on all AMPs. The size of Spool 3 is estimated to be 4
      rows. The estimated time for this step is 0.08 seconds.
3) We execute the following steps in parallel.
   a) We do an all-AMPs JOIN step from Spool 3 (Last Use) by way of an
      all-rows scan, which is joined to TEST.Size. Spool 3 and
      TEST.Size are joined using a product join. The result goes into
      Spool 4, which is built locally on the AMPs. The size of Spool 4
      is estimated to be 20 rows. The estimated time for this step is
      0.23 seconds.
   b) We do an all-AMPs RETRIEVE step from Spool 2 (Last Use) by way of
      an all-rows scan into Spool 5, which is duplicated on all AMPs.
      The size of Spool 5 is estimated to be 20 rows. The estimated time
      for this step is 0.27 seconds.
4) We do an all-AMPs JOIN step from Spool 4 (Last Use) by way of an
   all-rows scan, which is joined to Spool 5 (Last Use). Spool 4 and
   Spool 5 are joined using a product join. The result goes into Spool
   6, which is duplicated on all AMPs. The size of Spool 6 is estimated
   to be 400 rows. The estimated time for this step is 4.19 seconds.
5) We do an all-AMPs JOIN step from Spool 6 (Last Use) by way of an
   all-rows scan, which is joined to TEST.Widgets by way of index # 4.
   Spool 6 and TEST.Widgets are joined using a nested join. The result
   goes into Spool 1, which is built locally on the AMPs. The size of
   Spool 1 is estimated to be 200 rows. The estimated time for this
   step is 22.73 seconds.
6) Finally, we send out an END TRANSACTION step to all AMPs involved in
   processing the request.
-> The contents of Spool 1 are sent back to the user as the result of
   statement 1. The total estimated time is 27.40 seconds.