Date: Tue, 15 Oct 2002 12:27:53 +0100
Reply-To: Peter Crawford <peter.crawford@DB.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Crawford <peter.crawford@DB.COM>
Subject: Re: SAS is slow? (123 mb/sec on pc???)
Content-type: text/plain; charset=iso-8859-1
Like "Mike Rhoads <RHOADSM1@WESTAT.COM>" says about
options dsoptions=profile nocenter;
"That certainly IS interesting -- haven't heard of that one before"
This seems to be a global way of setting data statement
options like /debug
try
options dsoptions=debug ;
data _null_;
do x = 1 to 1e4;
y = x * 2;
z = x * 3;
end;
run;
Interestingly, it seems dificult to switch off dsoptions=profile
option dsoptions=noprofile;
had no effect except
NOTE: The DATA step option noprofile is unknown.
Fortunately,
option dsoptions=normal;
had the desired effect , even though
NOTE: The DATA step option normal is unknown.
but the cpu time returned to "normal"
I begin to wonder about other datastep options
1 DATA /?;RUN;
-
22
76
ERROR 22-322: Syntax error, expecting one of the following: BUFFERED,
NOPASSTHRU, PASSTHRU, PGM, UNBUFFERED, VIEW.
ERROR 76-322: Syntax error, statement will be ignored.
NOTE: The SAS System stopped processing this step because of errors.
So that indicates only certain dsoptions are expected.
Notice that as well as profile, the "expecting" list
does not include DEBUG
Equally NOdebug is accepted but not documented.
Unfortunately my theories get confounded again when I
try profile as a datastep option in a data statement
12 data /profile;run;
-------
22
76
ERROR 22-322: Syntax error, expecting one of the follo......
So perhaps option dsoptions = xxxxxxxx is not a substitute for
data / xxxxxxxxx ;
HTH
Peter Crawford
Datum: 14/10/2002 21:05
An: SAS-L@LISTSERV.UGA.EDU
Antwort an: Mike Rhoads <RHOADSM1@WESTAT.COM>
Betreff: Re: SAS is slow? (123 mb/sec on pc???)
Nachrichtentext:
That certainly IS interesting -- haven't heard of that one before, but I see
it is mentioned if you run PROC OPTIONS INTERNAL. (Otherwise undocumented
as far as I can tell)
Mike Rhoads
Westat
RhoadsM1@Westat.com
-----Original Message-----
From: Jack Hamilton [mailto:JackHamilton@FIRSTHEALTH.COM]
Sent: Monday, October 14, 2002 12:00 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: SAS is slow? (123 mb/sec on pc???)
I don't know or use SCL, so I can't compare with the SCL profile, but
you can try
options dsoptions=profile;
which produces, for the following input, in SAS 8.2 under Windows NT:
-----
options dsoptions=profile nocenter;
data _null_;
do x = 1 to 1e6;
y = x * 2;
z = x * 3;
end;
run;
-----
this output:
-----
DATA Step Profile Statistics
CPU
Count Usage CPU % Phase
-------- -------- ------ --------------------
1 0.000 0.00 Initialization
1 0.000 0.00 Compilation
1 0.000 0.00 Resolution
1 0.000 0.00 Execution / Init
1 8.115 100.00 Execution
1 0.000 0.00 Execution / Term
-------- -------- ------ --------------------
Total CPU Time == 8.12
DATA Step Statement Profile Statistics
CPU Log
Count Usage CPU % Line# Source
-------- -------- ------ ------
-----------------------------------------
13 data _null_;
1 0.000 0.00 14 do x = 1 to 1e6;
1000000 2.503 30.84 15 y = x * 2;
1000000 2.747 33.85 16 z = x * 3;
1000000 2.865 35.30 17 end;
1 0.000 0.00 18 run;
-------- -------- ------ ------
-----------------------------------------
-----
--
JackHamilton@FirstHealth.com
Manager, Technical Development
METRICS Department, First Health
West Sacramento, California USA
>>> "Mauro Morandin" <second_name@LIBERO.IT> 10/13/2002 2:37 PM >>>
Hi Sigurd,
thanks for your help, I will surely post to this newsgroup well
defined
programming problems (not entire data models) in which I find that SAS
is slower than required.
This week I'm doing some profiling to see where we loose too much time
and will then look for help from you and the other SAS wizards here on
the newsgroup.
I was wondering if in SAS V8.2 or in SAS V9.0 there is a data step
profiler like the SCL Profiler????
I appreciate your help.
Ciao.
Mauro.
Sigurd Hermansen wrote:
> As for Oracle RAW devices, not all Oracle programmers recommend them.
See:
>
> http://www.geocrawler.com/archives/3/279/2001/1/0/5049827/
>
> Before considering specialized hardware and file access systems, show
us
> examples of the type of data warehousing data that are bogging down
SAS
> solutions. It should not take long to write a sAS program that will
generate
> adequate test data. Generated data do not have to be trivial. Some of
the
> examples that I have seen on SAS-L generate gigabytes of data. I
would bet
> that a combination of data reshaping and better SAS methods will put
SAS
> solutions at least on par with Oracle and SQLServer. Don't worry
about
> overwhelming us. I have witnessed more more than a few amazingly
efficient
> solutions to problems involving complex and voluminous data.
> Sig
>
> -----Original Message-----
> From: David L. Cassell [mailto:Cassell.David@EPAMAIL.EPA.GOV]
> Sent: Sunday, October 06, 2002 5:32 PM
> To: SAS-L@LISTSERV.UGA.EDU
> Subject: Re: SAS is slow? (123 mb/sec on pc???)
>
>
> Mauro Morandin <second_name@LIBERO.IT> replied (to me):
>
>>Have you tried Oracle on RAW DEVICES?
>>Have you tried MS SQL on RAID0 filesystems with large STRIPING BLOCK
>
> SIZE.
>
>>Probably you were using your standard Oracle installation (which is
>>optimized for transaction processing) to do data warehousing (which
is
>>large sequential read/write). Ask your DBA.
>
>
> Now this is valuable. (Not that your previous writings are not!)
> Thanks for the specifics. Have you tried either of these on a
system
> which is also optimized for SAS? (Other than the
less-than-cooperative
> AIX...) If you have any substantive benchmarks based on these,
> whether they make SAS look good or bad, they would be welcome in
this
> forum.
>
> David
> --
> David Cassell, CSC
> Cassell.David@epa.gov
> Senior computing specialist
> mathematical statistician
--
--
mauro morandin
SAS consultant
red hat certified engineer
mauro.morandin-NOSPAM-@ieee.org
--
matrix srl via postioma di salvarosa 25b tel +39-423-724620
31033 castelfranco veneto (tv) fax +39-423-770798
http://www.matrix-online.it
info-TAKEAWAY-@matrix-online.it
--
Diese E-Mail enthält vertrauliche und/oder rechtlich geschützte Informationen. Wenn Sie nicht der richtige Adressat sind oder diese E-Mail irrtümlich erhalten haben, informieren Sie bitte sofort den Absender und vernichten Sie diese Mail. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Mail ist nicht gestattet.
This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have received this e-mail in error) please notify the sender immediately and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly forbidden.
