sas, Running SAS on MacBook Pro with Vista OS

Dear all,

Just wondering if anyone has this setup.

MacBook Pro with Vista OS (bootcamp)

How fast does SAS run?

thank you,
claus

I don't have that exact setup, but if you're running under BootCamp,
not in Mac OS X with an emulator like VMWare, it should be the same
speed as similar hardware running Windows without BootCamp. And
that's probably more dependent on your disk speed, available RAM, and
CPU speed than it is on whether your machine has BootCamp installed.

If your files are in the Mac partition and you access them with
MacDrive, you might try moving them to a native Windows partition. I
haven't done any testing with MacDrive, but reading a Mac file from a
Windows VM using NFS is really slow.

There were some early reports of problems with large (2GB+) files
under BootCamp, but I think those have been resolved.

My paper "SAS on a Mac? Yes!" was rejected for SAS Global Forum, so
I'm unlikely to do much more benchmarking at this point.

Thank you Jack. I guess I have to give it more thoughts. Someone
mentioned to me that this setup is limited to smaller datasets because
of limited memory allocation (1 gigbytes in windows from Apple OS)

Have you tried running large datasets in SAS using this setup?

thanks,
josh

Hi, I've tried to run sas on MACBOOK both using VMWARE and Bootcamp. I
must say that I can't stand the performance. The machine gets hot and
responds slowly when I'm running big datasets. Since I need to deal
with datasets larger than 1Gb, I get a dell desktop to run sas and use
remote desktop connection to control the program. It works for me!

My MacBook gets hot reading email, so I can't blame SAS in particular
for that.

My guess is that your MacBook is just underpowered for what you are
trying to do. Notebooks in general have design constraints that make
them unsuitable for heavy computing applications - slower disks, less
memory, poor heat dissipation. It's not strictly a MacBook problem,
but Apple's design goals for notebooks (small, slick-looking machines)
make make them less suitable for these applications than larger,
clunkier-looking models from other manufacturers.

I'm not sure whether you meant a 1 GB memory limit or a 1 GB filesize
limit, but I tried running the following code in three environments:

=====
data test (compress=no);
length a1-a1000 $1000;
retain a1-a1000 ' ';
do i = 1 to 2100;
output;
end;
run;

data _null_;
set test;
run;
=====

This creates a single SAS data set slightly larger than 2GB.

Results:

Windows XP in VMWare Fusion:
=====
13 data test (compress=no);
14 length a1-a1000 $1000;
15 retain a1-a1000 ' ';
16 do i = 1 to 2100;
17 output;
18 end;
19 run;

NOTE: The data set WORK.TEST has 2100 observations and 1001 variables.
NOTE: DATA statement used (Total process time):
real time 3:40.04
cpu time 36.15 seconds


20
21 data _null_;
22 set test;
23 run;

NOTE: There were 2100 observations read from the data set WORK.TEST.
NOTE: DATA statement used (Total process time):
real time 1:21.60
cpu time 19.00 seconds
=====


Windows XP in VMWare Fusion BootCamp partition:
=====
1 data test (compress=no);
2 length a1-a1000 $1000;
3 retain a1-a1000 ' ';
4 do i = 1 to 2100;
5 output;
6 end;
7 run;

NOTE: The data set WORK.TEST has 2100 observations and 1001 variables.
NOTE: DATA statement used (Total process time):
real time 2:43.59
cpu time 43.57 seconds


8
9 data _null_;
10 set test;
11 run;

NOTE: There were 2100 observations read from the data set WORK.TEST.
NOTE: DATA statement used (Total process time):
real time 3:30.10
cpu time 26.14 seconds

=====

Finally, running in Windows XP natively:
=====
1 data test (compress=no);
2 length a1-a1000 $1000;
3 retain a1-a1000 ' ';
4 do i = 1 to 2100;
5 output;
6 end;
7 run;

NOTE: The data set WORK.TEST has 2100 observations and 1001 variables.
NOTE: DATA statement used (Total process time):
real time 1:27.25
cpu time 2.59 seconds


8
9 data _null_;
10 set test;
11 run;

NOTE: There were 2100 observations read from the data set WORK.TEST.
NOTE: DATA statement used (Total process time):
real time 1.59 seconds
cpu time 1.56 seconds
=====

Running natively is substantially faster, but not shockingly faster.
This minimal test completed under all environments.

n Nov 30, 7:33m, XXXX@XXXXX.COM (Jack Hamilton) wrote:
> retain a1-a1000 ' >;
> do i = 1 to 2>00;
> gt;utput;
>>end>
>
> dat> _null_;
> > set te>t;
> ru>; >> =====
>
> This creates a single SAS data set slightly large> t>an 2GB.
> >> >esults:
>
> Windows XP in VMW>re Fusio>:
> =====
> 13 data test (c>mpress=no);
> 14 ength a>-a1000 $1000;
> 15 eta>n a1-a1000 ' ';
> 16 >do i = 1 to 2100;> > 17 gt;output;
> >8 > nd;
> 19 run;
>
> NOTE: The data set WORK.TEST has 2100 observa>ions and 1001 variables.
> NOTE: DATA statement u>ed (Total process time):
> gt;eal time 3:40.04
>
>>20
> 21 >at> _null_;
> 22 et test;
> 23 run;
>
> NOTE: There were 2100 >bservations read from the data set WORK.TEST.
> N>TE: DATA statement used (Tota> process time):
> eal tim> gt;>:21.60
> pu time 9.00 seconds> > =====> >
> Windows XP in VMWare Fusi>n BootCamp partition:
> =====> > 1 ata test (compress=>o);
> 2 length a1->1000 $1000;
> 3 > retain >1-a1000 ' '>
> 5 utput;
> 6 end;
> 7 gt;un;
>
> NOTE: The data set WORK.TEST has 2100 ob>ervations and 1001 variables.> > NOTE: DATA statement used (Tota> p>oces> time):
> >al time >:43.59
> gt;> pu time 3.57 seconds
>
> 8
> 9 ata _null_;
> 10 > et test;
> 11 run;
>
> NOTE: There were >100 observations read from th> data set WORK.TEST.
> NOTE: DATA>st>tement u>ed>(Total process time):
> eal time gt;>:30.10
> pu time gt;6.14 seconds
>
> =====> >
> Finally, running in Wi>dows XP natively:
> ====>
> 1 ata tes> (compress=n>);
> 2 gt;>ength a1-a1000 $1000;
> 3 retain a1-a1000 ' ';
> 4 do i = 1>to 2100;
> 5 utput;
> 6 end;
> > un;
>
> NOTE: The data>set WORK.TEST has 2100 observatio>s >nd 1>01 variables.
> NO>E: DATA statement >sed (Total >ro>ess time):
> eal time 1:27.25
> pu time gt;.59 seconds
>
> 8
> 9 ata _null_;
> 11 run;
>> > NOTE: There were 2100 observat>ons read>fr>m the data set WORK.TEST.
> NOTE: DATA statement used (Total process t>me):
> eal time 1.59 seconds
> gt;>cpu t>me .>6 seconds
> =====
>
> Ru>ning natively is substantially faster, bu> n>t shockingly faster.
> This minimal test completed under >ll>e>vironments.
>
> --
> Jack Hamilton
> XXXX@XXXXX.COM >org
> Vi>e>>s illam spi<>re>l>>ertatis auram
>
> On Nov 29, 2008, at 12:>8 >m>> XXXX@XXXXX.COM wrote:
>
> > On No> 2>,>>:12 pm, " XXXX@XXXXX.COM >m"><>>ee8...@gmail.>o>> wrote: >> >>>Dear all,
>
> >> Just wondering if anyone has this setup.
>
> >> >a>Book Pro with Vista OS (bootcamp)
>
> >> How fast does SAS run?
>
> >> >hank you,
> >> claus
>
> > Thank you Jack. guess I have to give>it>m>re thoughts. omeone
> > mentioned to me that this setup is lim>te> >o smaller >a>asets because
> > of limited memory allocation (1 gigbytes in windows from Apple OS)

Thank you Jack for your runs. I really appreciate your help!

josh

ummary: SGF and SAS on a MAC
#iw-value=1

Jack,

In part you wrote:

My paper "SAS on a Mac? Yes!" was rejected for SAS Global Forum,
so I'm unlikely to do much more benchmarking at this point.

That is most unfortunate. In general, your talks are always
interesting, and I would not miss them when fortunate enough to have
the opportunity. In this particular case, given the subject and the
speaker, almost nothing short of failing to go to SGF would have
prevented me from attending.

What section refused the paper?

I asked the institute support staff at SESUG about your subject and
got no practical information, which is rather unusual. Now I
understand that the institute has chosen not to support their product
when run on a non-windows operating system. But as I understand it,
running windows under boot camp is running a supported CPU on a
supported operating system in a native manner, so I find the lack of
response troubling. In particular, I wonder why some computer makers
are supported, but others are not, and how the decision is made.

What was the specific hardware used for the tests below?

Thanks.
Ian Whitlock

Date: Sun, 30 Nov 2008 19:33:28 -0800
Reply-To: Jack Hamilton < XXXX@XXXXX.COM >
Sender: "SAS(r) Discussion"
From: Jack Hamilton < XXXX@XXXXX.COM >
Subject: Re: Running SAS on MacBook Pro with Vista OS
Comments: To: " XXXX@XXXXX.COM " < XXXX@XXXXX.COM >
In-Reply-To: <0a18561a-1b71-4003-
XXXX@XXXXX.COM >
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes

I'm not sure whether you meant a 1 GB memory limit or a 1 GB filesize
limit, but I tried running the following code in three environments:

=====
data test (compress=no);
length a1-a1000 $1000;
retain a1-a1000 ' ';
do i = 1 to 2100;
output;
end;
run;

data _null_;
set test;
run;
=====

This creates a single SAS data set slightly larger than 2GB.

Results:

Windows XP in VMWare Fusion:
=====
13 data test (compress=no);
14 length a1-a1000 $1000;
15 retain a1-a1000 ' ';
16 do i = 1 to 2100;
17 output;
18 end;
19 run;


NOTE: The data set WORK.TEST has 2100 observations and 1001 variables.
NOTE: DATA statement used (Total process time):
real time 3:40.04
cpu time 36.15 seconds

20
21 data _null_;
22 set test;
23 run;


NOTE: There were 2100 observations read from the data set WORK.TEST.
NOTE: DATA statement used (Total process time):
real time 1:21.60
cpu time 19.00 seconds
=====

Windows XP in VMWare Fusion BootCamp partition:
=====
1 data test (compress=no);
2 length a1-a1000 $1000;
3 retain a1-a1000 ' ';
4 do i = 1 to 2100;
5 output;
6 end;
7 run;

NOTE: The data set WORK.TEST has 2100 observations and 1001 variables.
NOTE: DATA statement used (Total process time):
real time 2:43.59
cpu time 43.57 seconds

8
9 data _null_;
10 set test;
11 run;

NOTE: There were 2100 observations read from the data set WORK.TEST.
NOTE: DATA statement used (Total process time):
real time 3:30.10
cpu time 26.14 seconds

=====
Finally, running in Windows XP natively:
=====
1 data test (compress=no);
2 length a1-a1000 $1000;
3

uys,

1) From when I was at SAS, the decision of what platforms/hardware to
support was made on a very rough cost/benefit basis. Of course, currently
or previously supported platforms generally have a lower cost to maintain
and customers demanding continued support.

2) When you call SAS for support, just say you are on SAS 9.x.x running
Windows XP. If they can't replicate the problem on Windows XP, then they
would kick it back to you and your Max + XP environment. I doubt this will
happen very often.

3) Mac + XP + SAS is still a niche for their core corporate market- where
most of the revenue comes from. That said, there are definitely advocates
of adding this to the matrix at SAS, but again it all comes back to point 1
and enough customer demand.

4) Some vendors spend considerable sums of money and resources (support and
developers) to ensure their platform is supported by SAS, especially
higher-end servers.

5) There's always JMP, which started on the Mac and is on both the Mac and
Windows.

With hardware so cheap I would just kick in the $$$'s for a dedicated
Windows desktop and Remote Desktop from your Mac... I have had friends at
prior companies go both routes successfully, but decided it was simpler to
use the dual system route.

Regards,
Stephen

On Mon, Dec 1, 2008 at 1:59 PM, Ian Whitlock < XXXX@XXXXX.COM > wrote:

n Dec 1, 2008, at 2:42 pm, Stephen McDaniel wrote:


It would be very expensive to support SAS natively on a Mac - not that
the making it run would be so difficult, but verifying and supporting
it would take a lot of time and effort, and there's probably not a big
enough market to support it.



The audience for my paper might have been:

- Students who have Macs but not Windows machines, and

- Corporate employees who would like to work at home, and whose
corporate SAS license allows at-home use, but do not have Windows
machines at home, and

- Learners who have only the learning edition.

Yes, if you're spending someone else's money it might be better to
have a standalone Windows machine running SAS, but that's not
practical for many individuals.






On Dec 1, 2008, at 2:42 pm, Stephen McDaniel wrote: