Linux vs. Windows performance.
Michael Rothwell
configuration) and tested Oracle performance? I'm looking for some real
world examples if I move from Windows 2k to linux with Oracle 9i. The
box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
tables using about 30 G of space. About 40 - 100 concurrent users, but
this may double soon. Can I expect better performance from the Linux box?
Thanks
Michael
--
There is no worse tyranny than to force a man to pay for what he does
not want merely because you think it would be good for him. -- Robert
Heinlein
Howard J. Rogers
"Michael Rothwell" <marothwell...@yahoo.com> wrote in message
news:GeqdnbRkYfs...@comcast.com...
> Has anyone taken an Identical Windows box and Linux box (same hardware
> configuration) and tested Oracle performance? I'm looking for some real
> world examples if I move from Windows 2k to linux with Oracle 9i. The
> box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
> tables using about 30 G of space. About 40 - 100 concurrent users, but
> this may double soon. Can I expect better performance from the Linux box?
Yes. It's not real-world testing, but it's testing nonetheless. First time I
reformatted a Windows machine and turned it into Linux, I had the surprise
of my life. I had to re-run all my Oracle tests in case I had been measuring
Windows not Oracle, the difference was that dramatic. I can't quantify it
for you (well, I can say test were running at least 2, 3 and 4 times
faster -identical machine, I remind you- on Linux than on Windows, but I
didn't document that, and as I say, these were private stress tests, not a
production run), but yes you will notice it.
Group regulars will affirm, I think, that I was once a dedicated Linux
hater. Now I am merely incompetent on it. But despite that, I refuse to run
Oracle on Windows any more. It was that much of a shock.
Regards
HJR
--
--------------------------------------------
Oracle Insights: www.dizwell.com
--------------------------------------------
Daniel Morgan
> Has anyone taken an Identical Windows box and Linux box (same hardware
> configuration) and tested Oracle performance? I'm looking for some real
> world examples if I move from Windows 2k to linux with Oracle 9i. The
> box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
> tables using about 30 G of space. About 40 - 100 concurrent users, but
> this may double soon. Can I expect better performance from the Linux box?
>
> Thanks
>
> Michael
Some but how much may be dependent upon many things including the
flavour of Linux chosen. Keep in mind that changing the O/S is not
going to speed up the spin rate on the hard disks or bus speed or
CPU speed.
What I would expect to see is greater stability and far fewer
concerns with viruses and other mischief. Also consider the money
saved in not playing the MS upgrade game.
--
Daniel Morgan
http://www.outreach.washington.edu/ext/certificates/oad/oad_crs.asp
http://www.outreach.washington.edu/ext/certificates/aoa/aoa_crs.asp
damo...@x.washington.edu
(replace 'x' with a 'u' to reply)
Frank van Bortel
> Has anyone taken an Identical Windows box and Linux box (same hardware
> configuration) and tested Oracle performance? I'm looking for some real
> world examples if I move from Windows 2k to linux with Oracle 9i. The
> box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
> tables using about 30 G of space. About 40 - 100 concurrent users, but
> this may double soon. Can I expect better performance from the Linux box?
>
> Thanks
>
> Michael
>
Yes, I have - and taken OUT memory while at the process (512M->384M)
Switched from W2K (which once started as NT4) to RH7.3.
Database involved was an 8iEE Rel. 3 (8.1.7), used as development
machine; 2 433MHz Celerons, 2 IDE HD's (different size, though).
It died about 6 months ago - the MoBo is still on my desk.
Impressions (no measured tests, it as not expected!): on Linux,
about 20-30% faster (DB Creation time is a fine example, as all
my own scripts usually have the same options and parameters).
Not even an optimized kernel; which is actually a must!
--
Regards,
Frank van Bortel
Rick Denoire
>Has anyone taken an Identical Windows box and Linux box (same hardware
>configuration) and tested Oracle performance? I'm looking for some real
>world examples if I move from Windows 2k to linux with Oracle 9i. The
>box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
>tables using about 30 G of space. About 40 - 100 concurrent users, but
>this may double soon. Can I expect better performance from the Linux box?
Someone will perhaps confirm or deny these differences: There is no
such thing like shared memory in Windows (disadvantage). The lack of
OS buffering in Windows turns out to be a "built-in" advantage when
running Oracle. In most cases, one would rather mount file systems
with the directio option to avoid double buffering. Windows does not
need to disable what it does not have.
Whenever I ran Oracle on Windows and Linux, these were different
machines, so I can't really compare. But I have only seen Oracle
getting unrestartable only in Windows, and exactly only once.
Bye
Rick Denoire
Paul Drake
> configuration) and tested Oracle performance? I'm looking for some real
> world examples if I move from Windows 2k to linux with Oracle 9i. The
> box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
> tables using about 30 G of space. About 40 - 100 concurrent users, but
> this may double soon. Can I expect better performance from the Linux box?
>
> Thanks
>
> Michael
I have a Dell PE 2650 dual 3.06 GHz 512 KB L2 cache 1 MB L3 cache box
that is a dual boot of w2k3 server, RH9 server that I was going to
perform benchmarks on.
haven't yet. Will try to do so next week.
suggestions for test material welcomed.
Pd
Paul Drake
Rick,
lack of shared memory is not a disadvantage in this case, as it is not
needed for win32.
It is just different.
There is a single OS process per oracle instance.
As there are not multiple OS processes, why would one need to share
memory across them?
Please consult the docs on Metalink for this.
This is a solved problem, a non-issue. It is well documented, and just
requires you reading the docs.
Pd
<ducking and running>
why is there so much fuss about threading out there these days, if
multiple threads under a process are so evil?
NorwoodThree
On the Linux side, I know that if you use Dell equipment and a Pentium
4 processor, you can enable hyper-threading for the processor in the
BIOS. This essentially gives you 2 processors in one. It basically
is a virtual 2 processor configuration.
Not sure if Windows takes advantages of the HyperThreading, but Linux
and Oracle do. And it is fully supported.
Paul Drake
"supported" is a funny term here.
for w2k server, if there is a single physical CPU installed with
hyperthreading enabled, 2 CPUs will be recognized by the OS, which
will not be able to differentiate between physical and logical
processors.
for w2k3 server, if there is a single physical CPU installed with
hyperthreading enabled, 2 CPUs will be recognized by the OS, which
will be able to differentiate between physical and logical processors.
if the OS were to schedule execution of 2 processes on both the
physical and logical CPUs of a single physical processor, while not
schedule execution of a process on either the physical or logical CPU
of the second physical processor, I would certainly not consider that
to be "support".
this has been covered in detail in this forum before, as well as in
documentation available at Microsoft's knowledgebase.
Unfortunately, I can't help you out with benchmarks on such things for
now.
Pd
Toby Brown
Is hyper-threading really a performance advantage?
I was under the impression that hyper-threading was more of a
concurrency advantage (a clever technique to provide better
concurrency). I mean, taking your above example, you really only have
1 physical processor, therfore all processes will execute on it
serially, regardless of whether the O/S can be "fooled" into believing
that it is scheduling on 2 processors.
Or am I missing something about hyper-threading and it's performance
implications?
Toby Brown
Norwood, my apologies. There is actually a performance advantage
associated with Hyper-threading. I had been misinformed before.
Rick Denoire
>if the OS were to schedule execution of 2 processes on both the
>physical and logical CPUs of a single physical processor, while not
>schedule execution of a process on either the physical or logical CPU
>of the second physical processor, I would certainly not consider that
>to be "support".
And consider CPU affinity for processes being put in an out of the
queue. If the OS is not aware that the logical CPU resides in the same
physical CPU whose cache holds process data, the cached data will be
invalidated I think.
I use the same PE2550 (2 CPUs) with Redhat Linux AS 2.1 and Oracle
9.2.0.4. At times, CPU load is short of 400%!
By the way, I consulted Oracle on this question and they did not find
anything against hyperthreading in the specs, I was told.
Bye
Rick Denoire
Bob Jones
"Toby Brown" <toby_...@optusnet.com.au> wrote in message
news:2a84f50c.04021...@posting.google.com...
You have not been completely misinformed. It depends on how you look at it.
Hyperthreading does increase CPU utilization but not overall throughput. If
CPUs are 100% busy, hyperthreading will not do any good. Advertising one CPU
as two is misleading unless the key words "physical" and "logical" are
mentioned.
Niall Litchfield
news:1ac7c7b3.04021...@posting.google.com...
> <ducking and running>
> why is there so much fuss about threading out there these days, if
> multiple threads under a process are so evil?
Hi Paul
I believe there is some effort going into multi-threading on Linux, at which
point in accordance with standard IT practice it will become the best thing
since sliced bread.
--
Niall Litchfield
Oracle DBA
Audit Commission UK
NorwoodThree
Michael Rothwell
Thanks, this is the kind of information that I was looking for. We have
an instance that we are considering moving the OS to linux, and just
wanted some idea of what to expect with performance.
Ernest Siu
the ALU and the FPU. Therefore, if operations on the queue has both
logical operations and floating-point operations, they will be
parallelized. I'm not an Oracle expert or anything close to that but
I'll doubt a database query will require a lot of 'floating-point'
operations.
The BIOS says there're 2 CPU but there's really 1 physical CPU. Yeah
really :)
Ernest
norwoo...@my-deja.com (NorwoodThree) wrote in message news:<ba03e2c.04021...@posting.google.com>...
> Keyword: "Virtual"
Paul Drake
> NorwoodThree wrote:
> > Something else to consider-
> >
> > On the Linux side, I know that if you use Dell equipment and a Pentium
> > 4 processor, you can enable hyper-threading for the processor in the
> > BIOS. This essentially gives you 2 processors in one. It basically
> > is a virtual 2 processor configuration.
> >
> > Not sure if Windows takes advantages of the HyperThreading, but Linux
> > and Oracle do. And it is fully supported.
>
> Thanks, this is the kind of information that I was looking for. We have
> an instance that we are considering moving the OS to linux, and just
> wanted some idea of what to expect with performance.
>
> Michael
last time I checked, the 2.4 kernel did not differentiate between
physical and logical CPUs. That functionality was in 2.5 (dev) and
would be in 2.6.
someone may have backported that functionality into a 2.4 kernel.
In any event, disabling Hyperthreading is always an option, but a
reboot is required. Testing would be best.
I believe that OSDL (open source development labs) did some testing of
Hyperthreading.
check:
http://www.osdl.org/cgi-bin/eidetic.cgi?command=display&modulename=projects&on=1149
"Identify Hyper-threading performance improvements in Linux 2.4 and
2.5"
and this is worth a read:
http://www-124.ibm.com/developerworks/oss/linuxperf/
a quick search of the web returned this off of the Debian mailing list
http://lists.debian.org/debian-user/2003/debian-user-200301/msg00720.html
So - if you are still running a 2.4.x kernel, I'd disable it if you
have more than 1 physical CPU. When you are ready to install a 2.6.x
kernel, re-enable it.
Oracle was very non-specific regarding hypterthreading and per-CPU
licensing, so I had disabled it for servers that had Oracle software
installed. A month or so back, we received word that the per-CPU
licensing was per physical processor, not virtual processor, and that
hypterthreading _could_ be enabled.
as always, check with your Oracle Salesperson.
Pd
Chuck
> Has anyone taken an Identical Windows box and Linux box (same hardware
> configuration) and tested Oracle performance? I'm looking for some
> real world examples if I move from Windows 2k to linux with Oracle 9i.
> The box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP
> system with tables using about 30 G of space. About 40 - 100
> concurrent users, but this may double soon. Can I expect better
> performance from the Linux box?
>
> Thanks
>
> Michael
>
I have not run Oracle on Linux but I have on Windows, AIX, HPUX, IRIX,
Novell and SunOs over the past 9 years. 10% of my databases are on
Windows and they easily account for 90% of the problems - memory that
doesn't get released, process that get orphaned and must be manually
cleaned up, threads that get orphaned and cannot be cleaned up apart from
a reboot. I am looking forward to running Oracle on Linux perhaps in the
next year or two just as soon as I can convince management that Windows
is a crappy platform for Oracle.
The only problems I encounter with Oracle on Unix are those introduced by
application code.
--
Chuck
Remove "_nospam" to reply by email
John Gardner
In message <1076700909.575761@yasure>, Daniel Morgan
<damo...@x.washington.edu> wrote:
>
> Some but how much may be dependent upon many things including the
> flavour of Linux chosen. Keep in mind that changing the O/S is not
> going to speed up the spin rate on the hard disks or bus speed or
> CPU speed.
True, but such issues as Window's over aggressive paging system (ie it tries
to keep too many pages free to speed up application startup because that makes
gui programs look faster) is going to impact the beahviour of Oracle. It's
not always how fast your disk drive is, but what you do with it.
Regards
John G
Paul Drake
> Michael Rothwell wrote:
>
> > Has anyone taken an Identical Windows box and Linux box (same hardware
> > configuration) and tested Oracle performance? I'm looking for some real
> > world examples if I move from Windows 2k to linux with Oracle 9i. The
> > box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
> > tables using about 30 G of space. About 40 - 100 concurrent users, but
> > this may double soon. Can I expect better performance from the Linux box?
> >
> > Thanks
> >
> > Michael
>
> Some but how much may be dependent upon many things including the
> flavour of Linux chosen. Keep in mind that changing the O/S is not
> going to speed up the spin rate on the hard disks or bus speed or
> CPU speed.
>
> What I would expect to see is greater stability and far fewer
> concerns with viruses and other mischief. Also consider the money
> saved in not playing the MS upgrade game.
you're fooling yourself. you're still going to have to pay for the
operating system and support, and you're still going to have to
upgrade.
hey, I'd much rather send the funds to SuSE or Red Hat than to
redmond.
If you want to run Oracle on Linux _in production_ you're going to run
on a supported kernel and distribution and that is not free as in
beer.
you may be able to get by with whiteboxlinux. I didn't try it out, as
we picked up a development license for RHEL. It was quite reasonable.
Its more like $749/year for Red Hat Enterprise Linux 3.0 ES with
support.
Now, this is still less than support for w2k3 Server from an OEM like
HP.
Maybe you're never had to open a TAR. Congrats.
For the rest of us that depend upon Oracle Support to assist us when
the ORA-00600 messages are populating our logs and trace files, a
supported version is an operational requirement.
you're still going to be upgrading kernel versions, oracle versions,
applying patchsets (9.2.0.4 patchset 3, anyone?). how soon are you
going to hop on the 2.6.x train, making local stops at 2.6.3, 2.6.4
... 2.6.20?
one category you can put in your ROI, is, tripwire is free as in beer
on Linux.
that's roughly $500/server on win32.
Pd
Paul Drake
If I could ask for one feature to have working for Oracle on win32, it
would be dead connection detection. It was advertised to work in the
9.2.0.4 patchset, but did not. Maybe there will be a silver lining in
9.2.0.4 patchset 3 and DCD will work there. Maybe it will be in
9.2.0.5, or 10.1.0.2.0.
Or maybe we'll see such a rash of vulnerabilities due to the source
code release of a certain operating system that oracle on win32 will
be abandoned. :)
Pd
Chuck
news:1ac7c7b3.04021...@posting.google.com:
It was supposed to work on 8.1.5, 8.1.6 (what a horrible release), and
8.1.7 as well. In fact, way back on 8.1.5 (2 years ago maybe) I was told
by Oracle Tech.Support to enable DCD and that would fix my orphaned
threads problem. It did not. About a month ago, they told me it was the
cause of my orphaned threads problem, disable it and that would fix the
same problem. It did not. I wish I could simply move all these Win2k
instances to unix but the vendor will not support the application on any
platform but win2k. :-(
Frank van Bortel
Dump vendor - os has nothing to do with application that uses the db.
Hans Forbrich
> It was supposed to work on 8.1.5, 8.1.6 (what a horrible release), and
> 8.1.7 as well. In fact, way back on 8.1.5 (2 years ago maybe) I was told
> by Oracle Tech.Support to enable DCD and that would fix my orphaned
> threads problem. It did not. About a month ago, they told me it was the
> cause of my orphaned threads problem, disable it and that would fix the
> same problem. It did not. I wish I could simply move all these Win2k
> instances to unix but the vendor will not support the application on any
> platform but win2k. :-(
Unless the vendor has deliberatly put server-side hooks to the OS
(rare), there is no technical reason for running the database on any
specific OS.
You could ask the vendor to document why they need you to stay on
Windows, esp. since you are getting database support from Oracle.
One thought process to help deal with the vendor:
Are they running the app on the same box as the DB?
- If yes, are you running a multi-CPU machine?
- If yes, could you reduce the number of CPUs needed for the DB by
having the app on a separate machine?
- If yes, then have you licensed the DB by CPU?
- If yes, then when is the vendor going to reimburse you for the extra,
unnecessary DB licenses?
(does the phrase 'class action suit' have any meaning? <g>)
/Hans
Daniel Morgan
>>What I would expect to see is greater stability and far fewer
>>concerns with viruses and other mischief. Also consider the money
>>saved in not playing the MS upgrade game.
>
>
> you're fooling yourself. you're still going to have to pay for the
> operating system and support, and you're still going to have to
> upgrade.
You've lost me. Nowhere did I say Linux was free? I said "money saved
in not playing the MS upgrade game." Saving money means paying less
... not paying zero.
Ed Stevens
<fvanb...@netscape.net> wrote:
<snip>
>> I wish I could simply move all these Win2k
>> instances to unix but the vendor will not support the application on any
>> platform but win2k. :-(
>
>Dump vendor - os has nothing to do with application that uses the db.
Agreed. Apps shouldn't know anything about the os at the other end of
that pipe we call SQL*Net.
However, as we all know all too well, when the business customer has
already committed $$ to an app that he insists is absolutely vital to
his operation, it's nigh unto impossible to convince him that he spent
his money on a piece of s***.
Which is an entirely different problem from convincing a clueless app
vendor that, no, their app does NOT care which OS the DB is on.
Paul Drake
> Paul Drake wrote:
>
> >>What I would expect to see is greater stability and far fewer
> >>concerns with viruses and other mischief. Also consider the money
> >>saved in not playing the MS upgrade game.
> >
> >
> > you're fooling yourself. you're still going to have to pay for the
> > operating system and support, and you're still going to have to
> > upgrade.
>
> You've lost me. Nowhere did I say Linux was free? I said "money saved
> in not playing the MS upgrade game." Saving money means paying less
> ... not paying zero.
agreed.
Pd
Joel Garry
> For what I understand, hyperthreading is about concurrently utilizing
> the ALU and the FPU. Therefore, if operations on the queue has both
> logical operations and floating-point operations, they will be
> parallelized. I'm not an Oracle expert or anything close to that but
> I'll doubt a database query will require a lot of 'floating-point'
> operations.
Actually, Oracle spends quite a bit of time figuring out if things are
in memory [SGA] so it doesn't have to go plead for them from the disk.
This winds up being a big win on an OLTP system, as well as in use of
packaged procedures. One query doesn't make much difference, but a
whole bunch of one queries requires a bigger whole bunch of
operations. So the question in my mind becomes, is Oracle smart
enough to use hyperthreading if it is enabled to help out with this
stuff. The only answer to this I've seen appears to be "no, it just
uses the cpu count to set some minor assumptions."
>
> The BIOS says there're 2 CPU but there's really 1 physical CPU. Yeah
> really :)
>
> Ernest
>
> norwoo...@my-deja.com (NorwoodThree) wrote in message news:<ba03e2c.04021...@posting.google.com>...
> > Keyword: "Virtual"
jg
--
@home.com is bogus.
"Love is a perky elf dancing a merry little jig and then suddenly he
turns on you with a miniature machine gun." - Matt Groening, who
turned the big 50 the day after Valentine's day.
Chuck
news:c158tl$cvh$1...@news1.tilbu1.nb.home.nl:
> Chuck wrote:
>> I wish I could simply move
>> all these Win2k instances to unix but the vendor will not support the
>> application on any platform but win2k. :-(
>
> Dump vendor - os has nothing to do with application that uses the db.
Parts of the application run on the database server (purges, maintenance,
etc.) so in this case the db platform does have a bearing on the
application.
Chuck
news:oWpZb.28148$n17.25302@clgrps13:
>
>> It was supposed to work on 8.1.5, 8.1.6 (what a horrible release),
>> and 8.1.7 as well. In fact, way back on 8.1.5 (2 years ago maybe) I
>> was told by Oracle Tech.Support to enable DCD and that would fix my
>> orphaned threads problem. It did not. About a month ago, they told me
>> it was the cause of my orphaned threads problem, disable it and that
>> would fix the same problem. It did not. I wish I could simply move
>> all these Win2k instances to unix but the vendor will not support the
>> application on any platform but win2k. :-(
>
> Unless the vendor has deliberatly put server-side hooks to the OS
> (rare), there is no technical reason for running the database on any
> specific OS.
Automated maintenance jobs run on the DB server. Also, the vendor is
responsible for schema maintenace during upgrades. All of their upgrade
scripts are windows batch files and all of their technical expertise is
geared towards Windows.
I am trying to convince them to let me move the database to AIX, and they
can still run their maintenance/upgrades from the Win2k server. They
should be able to just set the LOCAL environment variable to the tns name
I give them for their database and everything should work. They have
never done this before however and are reluctant.
Frank van Bortel
Chuck, I'd say you were hired for the maintenance part (no offense),
and purges? Purging data from the database? That should have nothing
to do with the os platform, either.
Of course there are always arguments - from the vendor, mostly.
Nevertheless, I stand by my remark, although I would inform
management in somewhat other wording ;-)
Frank van Bortel
> Hans Forbrich <hfor...@yahoo.net> wrote in
> news:oWpZb.28148$n17.25302@clgrps13:
>
>
>>>It was supposed to work on 8.1.5, 8.1.6 (what a horrible release),
>>>and 8.1.7 as well. In fact, way back on 8.1.5 (2 years ago maybe) I
>>>was told by Oracle Tech.Support to enable DCD and that would fix my
>>>orphaned threads problem. It did not. About a month ago, they told me
>>>it was the cause of my orphaned threads problem, disable it and that
>>>would fix the same problem. It did not. I wish I could simply move
>>>all these Win2k instances to unix but the vendor will not support the
>>>application on any platform but win2k. :-(
>>
>>Unless the vendor has deliberatly put server-side hooks to the OS
>>(rare), there is no technical reason for running the database on any
>>specific OS.
>
>
> Automated maintenance jobs run on the DB server. Also, the vendor is
> responsible for schema maintenace during upgrades. All of their upgrade
> scripts are windows batch files and all of their technical expertise is
> geared towards Windows.
Which makes me shiver: their expertise should be Oracle, then Windows...
>
> I am trying to convince them to let me move the database to AIX, and they
> can still run their maintenance/upgrades from the Win2k server. They
> should be able to just set the LOCAL environment variable to the tns name
> I give them for their database and everything should work. They have
> never done this before however and are reluctant.
See - the above already gets proven; their knowledge of Oracle is
less-than-desired, to say it politely.
kenneth_koenraadt
I took 2 identical hardware configurations, 2 identical 9.2 DB's with
equally-sized SGAs on W2K and Red hat Linux 8 respectively. The 2 O/S'
had approximately the same dedicated ressources.
Then I created one million unique INSERT (which is quite
CPU-intensive) statements (with a Perl script) and ran them agains
each instance. Result : Linux was 150% faster than W2k.
- Kenneth Koenraadt
On Fri, 13 Feb 2004 11:25:40 -0800, Michael Rothwell
<marothwell...@yahoo.com> wrote:
>Has anyone taken an Identical Windows box and Linux box (same hardware
>configuration) and tested Oracle performance? I'm looking for some real
>world examples if I move from Windows 2k to linux with Oracle 9i. The
>box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
>tables using about 30 G of space. About 40 - 100 concurrent users, but
>this may double soon. Can I expect better performance from the Linux box?
>
>Thanks
>
>Michael
>
kenneth_koenraadt
I stand self-corrected :
Linux was 50% faster, not 150%.
- Kenneth Koenraadt
Rick Denoire
>Has anyone taken an Identical Windows box and Linux box (same hardware
>configuration) and tested Oracle performance? I'm looking for some real
>world examples if I move from Windows 2k to linux with Oracle 9i. The
>box currently has 2 1.8 CPUs with 4 disks and 1 G ram. OLTP system with
>tables using about 30 G of space. About 40 - 100 concurrent users, but
>this may double soon. Can I expect better performance from the Linux box?
Go to
http://www.interealm.com/technotes/roby/pentmark.html
Impressive comparison.
Bye
Rick Denoire
Joel Garry
I'm glad as heck someone finally published such a thing. A bit thin
in the exposition of methodology for sceptical replication though,
unless I missed something.
jg
--
@home.com is bogus.
http://www.signonsandiego.com/uniontrib/20040302/news_1b2adelphia.html
Michael Rothwell
Now that's a very good little site. Granted, no "in-depth", but he did
use the same hardware and same tests, so there is some good data.
Thanks.