Re: DRDA GATEWAY to IBM - SQL performance
Kate,
Its been about a year or so, but try this.
Run the query twice. (One right after the other.)
You should see the second one run faster.
If this is the case, then I might have an idea ....
-Mikey
Kate_Tomchik@homedepot.com wrote:
>
> Has anyone seen this before? Any suggestions on how we can improve the
> performance?
>
> ---------------------- Forwarded by Kate Tomchik/IS/SSC/THD on 05/13/98
> 12:08 PM ---------------------------
>
> I'm hoping you may be able to shed some light on a performance quirk I
> noticed when doing a join with a drda (remote) table. The goal is to
> populate a temp table with rows from a join of a local table and the remote
> table. rlctr_sku is local ans sku is remote. If you code the sql in a
> single statement:
>
> INSERT INTO temp_sku
> SELECT sku.sku_nbr, sku.sku_desc
> FROM rlctr_sku, sku
> WHERE rlctr_sku.sku_nbr = sku.sku_nbr>
> It takes about 3 minutes with the data I had in rlctr_sku.
>
> If I break this into an unload and a load:
>
> UNLOAD TO '/tmp/temp_sku'
> SELECT sku.sku_nbr, sku.sku_desc
> FROM rlctr_sku, sku
> WHERE rlctr_sku.sku_nbr = sku.sku_nbr;>
> LOAD FROM '/tmp/temp_sku'
> INSERT INTO temp_sku;>
> Then it takes about 40 seconds
>
> I did an sqlexplain on both and in each case they used the same path -
> index only on rlctr_sku, and the remote db was sent the same query in both
> cases. There was no information output about the INSERT efficiency.
--
#include <std_disclaimer.h> /* Mike Segel (MS385) */
#include <No_Spam.h>
#ifdef OFFENDED_BY_CONTENT
The author takes no responsibility for this post.
Any resembalence to a coherent rational thought is purely coincidence.
-The Management.
#endif
*****************************
* Attention *
-*- Due to AGIS's Refusal to Act Responsibly
-*- Due to ACSI's Refusal to Act Responsibly
We are blocking all of their domains at the packet level.
This block will exist until they modify their policies to
conform to existing RFCs and net community standards.
We encourage all ISPs and domain holders to do the same.
*****************************