Thread: performance problem with 3-column indexes

the table has about 30k records.    simple select statement, by primary key, requires plenty of cpu
time when the primary key has three columns   when the primary key has two columns several times less cpu is
required (even though the contents of the table is the same.

so:  PRIMARY KEY(C_ID, C_D_ID, C_W_ID) -> PRIMARY KEY (C_ID, C_D_ID)          select cpu: 600           ->
selectcpu 60
 


----------- how can i work around it? -----------------

and no, I can't blend the two columns into one, the spec does not
allow that - so is there some tuning parameter?


P.S. The table definition:


CREATE TABLE CUSTOMER (C_ID            SMALLINT NOT NULL,    /*--*/C_D_ID          SMALLINT NOT NULL,    /*--*/C_W_ID
      SMALLINT NOT NULL,   /*--*/C_FIRST         VARCHAR(16) NOT NULL,C_MIDDLE        VARCHAR(2) NOT NULL,C_LAST
 VARCHAR(16) NOT NULL,C_STREET_1      VARCHAR(20) NOT NULL,C_STREET_2      VARCHAR(20) NOT NULL,C_CITY
VARCHAR(20)NOT NULL,C_STATE         VARCHAR(2) NOT NULL,C_ZIP           INTEGER NOT NULL,    /*--*/C_PHONE
NUMERIC(16)NOT NULL,C_SINCE         timestamp,C_CREDIT        VARCHAR(2) NOT NULL,C_CREDIT_LIM    NUMERIC(12,2) NOT
NULL,C_DISCOUNT     NUMERIC(4,4) NOT NULL,C_BALANCE       NUMERIC(12,2) NOT NULL,C_YTD_PAYMENT   NUMERIC(12,2) NOT
NULL,C_PAYMENT_CNT  SMALLINT NOT NULL,    /*--*/C_DELIVERY_CNT  SMALLINT NOT NULL,    /*--*/C_DATA
VARCHAR(500)NOT NULL,PRIMARY KEY   (C_ID, C_D_ID, C_W_ID)
 
);

Can you give an example query and explain output for both cases?
Have you run vacuum analyze?

Since I haven't seen the query, one thing that might bite you would
be if you aren't casting your constants to smallint, although I
don't know why that would change on the index definition.

On 6 Nov 2001, charles wrote:

> the table has about 30k records.
>     simple select statement, by primary key, requires plenty of cpu
> time when the primary key has three columns
>     when the primary key has two columns several times less cpu is
> required (even though the contents of the table is the same.
>
> so:
>    PRIMARY KEY(C_ID, C_D_ID, C_W_ID) -> PRIMARY KEY (C_ID, C_D_ID)
>            select cpu: 600           ->        select cpu 60
>

czl@iname.com (charles) writes:
>     simple select statement, by primary key, requires plenty of cpu
> time when the primary key has three columns
>     when the primary key has two columns several times less cpu is
> required (even though the contents of the table is the same.

What does EXPLAIN show in the two cases?  What PG version is this?
        regards, tom lane