Date:         Mon, 11 Sep 2000 15:47:55 GMT
Reply-To:     amichiel@EARTHLING.NET
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         amichiel@EARTHLING.NET
Organization: Deja.com - Before you buy.
Subject:      Re: Disastrous performance in V8 cf V6

This 'thread' has been split into 2 pieces by my News software.
I just made a posting on the 'other' thread with a suggestion to
change the programming technique being used.

I missed this original posting information (or it had been cut out
of the thread when I saw it...).
You do know that the 'technique' you use to 'convert' the datasets
(tables) from V6 to V8 may impact (destroy) the index.  Now that I
read this, I suspect that the primary problem _is_ the index, or
rather, the lack of it.  If you simply used a datastep to create a
new V8 dataset (table), you no longer have a index, and need to use
a procedure to build the indexes.
In fact, we left out V6 files alone, and just converted to using V8
with the V6 data sources.


In article Stephen Dunn <Stephen_D@HWAY.CO.UK> wrote:
> We run a middling size SAS database (one large table
> of 10million rows, a few others of 100,000s rows) on an NT server
> which we have just converted to using V8 software and V8 table
> formats from V6. The database updates are now taking much longer
> than before and its performance on queries is much worse. The
> difficulty seems to be around accessing files in index order rather
> than physical order - in some cases this is taking almost ten times
> as long as V6 used to on the same data in V6 formats. Even
> the simple act of reading a file using an index as BY variables
> after a SET statement is painfully slow.
> Has anyone had similar experiences? If so, is it curable?
>                 Stephen Dunn
>                 Highway Insurance Statistician
>                 Tel  (01277 266) 253



Sent via Deja.com http://www.deja.com/
Before you buy.