Date: Thu, 22 Jan 2009 18:25:48 -0600
Reply-To: "./ ADD NAME=Data _null_;" <iebupdte@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "./ ADD NAME=Data _null_;" <iebupdte@GMAIL.COM>
Subject: Re: Work Folder ??
In-Reply-To: <b7a7fa630901221558i32e214f7w37c663b25ceafad9@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Try running the data steps as seperate BATCH program... no difference
at my house.
NOTE: SAS 9.1.3 Service Pack 4
NOTE: SAS initialization used:
real time 0.11 seconds
cpu time 0.12 seconds
1 options fullstimer=1;
2 DATA A;
3 INPUT A B;
4 do i=1 to 2000000;
5 output;
6 end;
7 cards;
NOTE: The data set WORK.A has 6000000 observations and 3 variables.
NOTE: DATA statement used (Total process time):
real time 19.99 seconds
user cpu time 0.38 seconds
system cpu time 0.53 seconds
Memory 144k
11 ;
NOTE: SAS Institute Inc., SAS Campus Drive, Cary, NC USA 27513-2414
NOTE: The SAS System used:
real time 20.10 seconds
user cpu time 0.43 seconds
system cpu time 0.60 seconds
Memory 1247k
NOTE: SAS 9.1.3 Service Pack 4
NOTE: SAS initialization used:
real time 0.12 seconds
cpu time 0.10 seconds
1 options fullstimer=1;
2 DATA work.A;
3 INPUT A B;
4 do i=1 to 2000000;
5 output;
6 end;
7 cards;
NOTE: The data set WORK.A has 6000000 observations and 3 variables.
NOTE: DATA statement used (Total process time):
real time 19.64 seconds
user cpu time 0.33 seconds
system cpu time 0.40 seconds
Memory 144k
11 ;
NOTE: SAS Institute Inc., SAS Campus Drive, Cary, NC USA 27513-2414
NOTE: The SAS System used:
real time 19.76 seconds
user cpu time 0.38 seconds
system cpu time 0.45 seconds
Memory 1247k
On 1/22/09, Joe Matise <snoopy369@gmail.com> wrote:
> I'm highly confused by the DATA A times. They're nearly identical to the
> CPU times. In my experience that's never true for writing out a significant
> dataset... I/O times are always far more significant than simple processing.
>
> Does that suggest that DATA A is writing to memory, and WORK.A is writing to
> disk?? I'm fairly sure that datasets for me are always written to disk ...
>
> I ran a similar test (identical code, except for correcting the missing
> CARDS; statement, and 1/4 as many records to make it not too slow on my
> desktop) and got much more variable results; this is on an early Core2Duo
> dual core machine (1GB RAM, SATA2 HDD)
> All runs are DATA A / DATA WORK.A and are actual times, not CPU times. CPU
> time is nearly identical across all runs, 0.50 sec +/- 0.05 sec.
>
> First run: 7.03 / 5.71
> Second: 14.26/15.62
> Third: 8.15/4.31
> Fourth: 3.42/3.76
>
> It's certainly possible that Outlook or something else might have started up
> during one or the other, so this is far from conclusive, but I certainly
> don't seem to see a difference here.
>
> Does your log show both like
> The data set WORK.A has 1500000 observations and 3 variables.
> (but with a different number of obs.)? Or does the one show something other
> than WORK.A ?
>
> -Joe
>
> On Thu, Jan 22, 2009 at 5:30 PM, Arthur Tabachneck <art297@netscape.net>wrote:
>
> > Hopefully, those more knowledgeable than me will chime in once I post this.
> >
> > I have always thought that the two were equivalent but, from the test I
> > just ran, they don't appear to be.
> >
> > The two programs I ran, fifteen times in succession, and their results are
> > shown below. I was extremely surprised that 'data a' ALWAYS ran faster
> > than 'data work.a'.
> >
> > Can anyone explain why I obtained those results? FWIW, the tests were run
> > on a 4-processor windows-based server running 9.1.3 after everyone else
> > had left the office.
> >
> > Art
> >
> > DATA A; DATA work.a;
> > INPUT A B; INPUT A B;
> > do i=1 to 2000000; do i=1 to 2000000;
> > output; output;
> > end; end;
> > cards;
> > 1 2 1 2
> > 1 3 1 3
> > 2 5 2 5
> > ; ;
> > results:
> >
> > real time 2.07 seconds 7.29 seconds
> > cpu time 2.01 seconds 2.01 seconds
> >
> > real time 2.06 seconds 5.87 seconds
> > cpu time 2.04 seconds 2.00 seconds
> >
> > real time 2.09 seconds 5.29 seconds
> > cpu time 2.07 seconds 2.01 seconds
> >
> > real time 2.17 seconds 6.10 seconds
> > cpu time 2.04 seconds 1.99 seconds
> >
> > real time 2.37 seconds 5.46 seconds
> > cpu time 2.04 seconds 2.03 seconds
> >
> > real time 2.09 seconds 4.73 seconds
> > cpu time 2.09 seconds 2.00 seconds
> >
> > real time 2.01 seconds 5.67 seconds
> > cpu time 1.99 seconds 2.10 seconds
> >
> > real time 2.07 seconds 5.38 seconds
> > cpu time 2.09 seconds 2.04 seconds
> >
> > real time 2.15 seconds 8.45 seconds
> > cpu time 2.00 seconds 2.01 seconds
> >
> > real time 2.01 seconds 4.53 seconds
> > cpu time 2.00 seconds 1.97 seconds
> >
> > real time 2.01 seconds 5.03 seconds
> > cpu time 2.00 seconds 2.01 seconds
> >
> > real time 2.04 seconds 6.10 seconds
> > cpu time 2.03 seconds 2.06 seconds
> >
> > real time 5.65 seconds 6.48 seconds
> > cpu time 2.04 seconds 2.04 seconds
> >
> > real time 2.17 seconds 5.62 seconds
> > cpu time 2.03 seconds 1.98 seconds
> >
> > real time 2.04 seconds 6.67 seconds
> > cpu time 2.01 seconds 2.01 seconds
> > --------
> > On Thu, 22 Jan 2009 16:28:50 -0500, SAS_learner <proccontents@GMAIL.COM>
> > wrote:
> >
> > >Hello _all_,
> > >
> > >Once upon a time Long ago one fellow SAS programer told me that there is
> > >difference between way we create work datasets. Say for example
> > >
> > >data test ;
> > > set sashelp.class;
> > >Run;
> > >
> > >is different
> > >
> > >Data work.test;
> > > set sashelp.class ;
> > >run;
> > >
> > >In the Second Method dataset test is stored in the work folder (either on
> > C:
> > >or Network ) ?? When You delete where would those datasets go not to
> > recycle
> > >bin
> > >
> > >thanks
> > >SL
> >
>
| 
