Forcing NT to use RAM for PAGEFILE.SYS using RAMDISK ??
Michael Seamans
I was playing around with the ramdisk utility I found for
NT (pretty nice...the driver is only 25K in size & can be
set up in less than 5 min.) and had this idea. Could you
run NT with all of your PAGEFILE.SYS file on a ramdisk??
Currently, I am experimenting with ramdisk & I'm going to
try this theory out. In theory, you should be able to
create a ramdisk that could be large enough to hold your
pagefile & every time you boot, your ramdisk is created
(the driver boots at the BOOT level in NT, so it is
available when all of your SYSTEM & above drivers start),
NT creates your pagefile on your ramdisk & it is used for
all paging. Now to do this you would probably need at least
64-80M of system memory (40M for NT and 24-40M for
pagefile), but the performance would be outstanding (in
theory). My question is this - has anybody out there tried
this??? If so, what were your results??
--
Michael Seamans
SEA...@NOSPAM.EROLS.COM
If replying, drop the NOSPAM from my address
"Some people see things as they are & ask 'Why'; I dream of
things yet to be & say 'Why not'" - George Bernard Shaw
The NTGURU
Michael, I think that with 80 meg of RAM you will find that it runs
plenty fast! NT is very smart about it's memory usage and uses whatever
is not in current use for a disk cache. I don't think you will see a
performance increase. I also don't think that you will get it to work.
NT needs the virtual memory file during the boot process. Unless your
RAMDRIVE can be loaded at boot time and set to run following NTFS.SYS I
don't think you will have any luck...
Steve Shaw
Michael Seamans wrote in article <01bc84ef$b049d490$1d47accf@developer>...
>I was playing around with the ramdisk utility I found for
>NT (pretty nice...the driver is only 25K in size & can be
>set up in less than 5 min.) and had this idea. Could you
>run NT with all of your PAGEFILE.SYS file on a ramdisk??
>Currently, I am experimenting with ramdisk & I'm going to
>try this theory out. In theory, you should be able to
>create a ramdisk that could be large enough to hold your
>pagefile & every time you boot, your ramdisk is created
>(the driver boots at the BOOT level in NT, so it is
>available when all of your SYSTEM & above drivers start),
>NT creates your pagefile on your ramdisk & it is used for
>all paging. Now to do this you would probably need at least
>64-80M of system memory (40M for NT and 24-40M for
>pagefile), but the performance would be outstanding (in
>theory). My question is this - has anybody out there tried
>this??? If so, what were your results??
>---------
Great idea -- NT Workstation could use some help in the memory management
department. Let's think about this.... Suppose you have 50 meg of RAM;
NT's default pagefile size is RAM size plus 12, or 62 meg. Might be a
tight squeeze getting that pagefile into 50 meg of RAM. And that's just
for starters -- are you planning to run some apps also?
Still, it's a great idea! Keep 'em coming! :)
Steve Shaw
Larry Bernstone
What's the point? When NT initializes, it clears out all that RAM space
and uses it preferentially over disk space. If you run NT on a very large
RAM space (over 128MB), there is little access to the pagefile. What it
does put there is small bits of stuff that it figures won't be needed very
much. The extra overhead that this would cause in a RAMdisk (sectors are
4K, right?) would far outweigh any benefits. This is all IMHO, and I am no
NT expert, but I think those guys at MS have a pretty good idea what they
are doing on this sort of thing...Let the Ph.D.s do their work...
Michael Seamans wrote in article <01bc84ef$b049d490$1d47accf@developer>...
>I was playing around with the ramdisk utility I found for
>NT (pretty nice...the driver is only 25K in size & can be
lots cut....
ls0916
Hi Michael,
Why on earth would you want to put a Pagefile in a Ramdisk?
You're completely missing the point of what a pagefile is
used for.
A pagefile is used because your system needs more memory
than it physically has, therefore it swaps out some memory
temporarily to disk.
When you create a 40 MB ramdisk, you're decreasing the amount
of memory available to all your programs, and you'll be
increasing the amount of swapping, and kill performance.
You're better off just getting rid of the ramdisk altogether,
and increasing the total amount of physical memory to the system,
instead of making your system do somersaults trying to get at
your pagefile. Increase your physical memory enough, and you'll
experience so little paging that the bottleneck in performance
will be somewhere else (probably you're I/O system).
Steve
Michael Seamans wrote:
[snip]
> run NT with all of your PAGEFILE.SYS file on a ramdisk??
> Currently, I am experimenting with ramdisk & I'm going to
> try this theory out. In theory, you should be able to
> create a ramdisk that could be large enough to hold your
> pagefile & every time you boot, your ramdisk is created
> (the driver boots at the BOOT level in NT, so it is
> available when all of your SYSTEM & above drivers start),
> NT creates your pagefile on your ramdisk & it is used for
> all paging. Now to do this you would probably need at least
> 64-80M of system memory (40M for NT and 24-40M for
> pagefile), but the performance would be outstanding (in
> theory). My question is this - has anybody out there tried
> this??? If so, what were your results??
>
> --
> Michael Seamans
> SEA...@NOSPAM.EROLS.COM
kru...@acsu.buffalo.edu
> Why on earth would you want to put a Pagefile in a Ramdisk?
> You're completely missing the point of what a pagefile is
> used for.
Exactly.
An easy to think about the Windows NT pagefile is like this:
Most computers have a Level 1 cache in the CPU and a Level 2
cache on the motherboard, then system ram. Think of a Windows NT
system as having L1 cache, L2 cache and "system ram" as Level 3
cache with the *hard drive pagefile* being the actual system
"memory".
A pagefile is a different thing from a SWAP file (as used in
Win95 and Win 3.x).
Roger
Steve Shaw
kru...@acsu.buffalo.edu wrote in article <33B7EB...@acsu.buffalo.edu>.
..
>An easy to think about the Windows NT pagefile is like this:
>
>Most computers have a Level 1 cache in the CPU and a Level 2
>cache on the motherboard, then system ram. Think of a Windows NT
>system as having L1 cache, L2 cache and "system ram" as Level 3
>cache with the *hard drive pagefile* being the actual system
>"memory".
>----------------------------
Interesting analogy, painfully true! And it's not a feature -- NT should
make better use of abundantly available RAM.
Steve Shaw
David Britton
soooo, have you tried it yet??? Pretty good idea... I would think that
around 80 megs (total memory) you might start seeing something happen
(positive)... you know it would be one of those "which curve grows faster"
problems. NT doesnt use memory well at all. I fail to see why MS has done
this to a very robust OS.... SOO much would be gained by keeping apps in
memory longer. I have 48 megs of ram on NTW, and it pages far too much. I
have Half my memory available, and still it pages to disk. If you could
access the disk less (the 1000x faster thing) benefits would be great i
would think...
in short: interesting idea... i would like to know what happens, and I too,
would like to know where you got that ramdisk utility. Contrary to
"popular" opinion on this thread, NT does a less than average job on memory
management. It really is one of its worst categories.
David
Michael Seamans <sea...@nospam.erols.com> wrote in article
<01bc84ef$b049d490$1d47accf@developer>...
> I was playing around with the ramdisk utility I found for
> NT (pretty nice...the driver is only 25K in size & can be
> set up in less than 5 min.) and had this idea. Could you
> run NT with all of your PAGEFILE.SYS file on a ramdisk??
> Currently, I am experimenting with ramdisk & I'm going to
> try this theory out. In theory, you should be able to
> create a ramdisk that could be large enough to hold your
> pagefile & every time you boot, your ramdisk is created
> (the driver boots at the BOOT level in NT, so it is
> available when all of your SYSTEM & above drivers start),
> NT creates your pagefile on your ramdisk & it is used for
> all paging. Now to do this you would probably need at least
> 64-80M of system memory (40M for NT and 24-40M for
> pagefile), but the performance would be outstanding (in
> theory). My question is this - has anybody out there tried
> this??? If so, what were your results??
>
> --
> Michael Seamans
> SEA...@NOSPAM.EROLS.COM
>
Dave Thomas
It is in fact a feature. EVERY multitasking OS does exactly the same
thing. I always see you complain about it, but that is what makes NT a
true multitasking system. If MS did what you suggest, NT would be just
like 95. If you don't like 95, why do you want NT to become 95? If NT
did not frequently trim working sets to disk, and maximize use of the
pagefile, then the system would not be prepared for another process to
start. The only way for an OS to maximize availability of resources is
to conserve them as much as possible. As you have been heard to lament,
this does not contribute to good foreground respponse time. However its
a tradeoff, so for it's high-end OS, MS chose on the side of maximum
multi-tasking ability, they aren't going to change it.
Dave
LEIBO
>It is in fact a feature. EVERY multitasking OS does exactly the same
>thing. I always see you complain about it, but that is what makes NT a
>true multitasking system.
EXACTLY right. If there is any problem here it's that MS hasnt really
given
us enough control to tune the performance behavior. This is probably
because:
- this is the first OS put out by MS where any real tuning could be done
- many NT users dont seem to understand the way a multitasking OS
works, so putting performance tweaking in their hands might be
dangerous!
kru...@acsu.buffalo.edu
> >cache with the *hard drive pagefile* being the actual system
> >"memory".
> Interesting analogy, painfully true! And it's not a feature -- NT should
> make better use of abundantly available RAM.
Well, it really isn't an "analogy"... it's how NT works. If you think
about it, you will see that as you add more physical memory, your
pagefile will actually get larger because the LRU (Least Recently
Used) algorithm will be swapping out more stale pages.
This method is used in almost all "real" operating systems (and
it works well).
I can see how a person could say "Hey I have 128 megs of RAM, what
the heck do I need a pagefile for?". Well, imagine that you could
install 64 or 128 megabytes of CACHE in your computer... why then
would you need DRAM? The answer? COST. Megabytes of static ram (fast
cache memory) costs a LOT more than DRAM. Likewise, megabytes of
hard drive space are cheaper than DRAM. So, why not let your hard
drive be your "system memory" and let the DRAM cache it for you?
It really does make sense once you think about it this way...
Roger
Steve Shaw
Dave Thomas wrote in article <33B7ED...@ml.com>...
>Steve Shaw wrote:
>>
>> kru...@acsu.buffalo.edu wrote in article <33B7EB...@acsu.buffalo.ed
u>.
>> ..
>> >An easy to think about the Windows NT pagefile is like this:
>> >
>> >Think of a Windows NT system as having L1 cache, L2 cache and "system
ram" as
>> Level 3 cache with the *hard drive pagefile* being the actual system
"memory".
>> >----------------------------
>>
>> Interesting analogy, painfully true! And it's not a feature -- NT
should
>> make better use of abundantly available RAM.
>>
>> Steve Shaw
>Steve,
>It is in fact a feature. EVERY multitasking OS does exactly the same
>thing.
Simply not true. High-performance multi-tasking O.S.'s make use of all
available memory instead of letting half of it sit idle. I'm not arguing
which OS is better, or that NT isn't a good O.S. I say that NT could be a
better, faster O.S. if it would use all it's memory, instead of half its
memory. (I'm talking NT Workstation here.)
>I always see you complain about it, but that is what makes NT a
>true multitasking system. If MS did what you suggest, NT would be just
>like 95. If you don't like 95, why do you want NT to become 95?
Actually, one of the good things about Win95 is that, unlike NT, Win 95
actually uses all the nice juicy RAM in a way that helps performance
instead of letting half (well OK, 40%) of it sit idle like NT Workstation
does. Maybe you're right that I've complained too much, but jeepers,
_somebody_ should tell Bill.....
>If NT did not frequently trim working sets to disk, and maximize use of
the
>pagefile, then the system would not be prepared for another process to
>start.
There is no reason to "frequently trim working sets to disk," unless RAM
is under load.
Keeping half the RAM unused simply to insure lots of free pages may make an
O.S. designer feel warm & fuzzy, but it slows down users who must then go
out and buy large amounts of RAM to run a system that fails to use RAM
efficiently. Please consider that performance comes from using RAM, not the
pagefile.
>The only way for an OS to maximize availability of resources is to
conserve them as much >as possible.
Is the goal to "maximize availability of resources?" This is easily done
-- just page half your RAM out to disk.
Wouldn't you really rather maximize performance?
>As you have been heard to lament, this does not contribute to good
foreground respponse >time.
Does paging contribute to good foreground response time?
For both Workstations and Servers, it makes sense to use most available RAM
to keep the most active tasks in memory. NT Workstation doesn't yet manage
to do this.
>However its a tradeoff, so for it's high-end OS, MS chose on the side of
maximum
>multi-tasking ability, they aren't going to change it.
They aren't? Rats. That leaves us with an NT that's firmly steeped in the
Pee Cee tradition, tuned as if "RAM is precious -- let's try not to use
it."
And of course NT can't be tuned. I complain about that sometimes too. :)
Still, NT remains my O.S. of choice. Actually, it's my second choice. My
first choice would be an NT lets me use all the machine I paid for.
Steve Shaw
AAD
There is an entry in the registry under memory management.
it has many subentries.
one will give option to choose between large or small file cache
soe of the other entries might be there to control the paging
system.
One entry of interest is disable paging executive
being the moderate NT4 user that I am I chose to ignore th entries and
and not mess with them.
some one with enough knowledge of the internals of NT might have a
better idea as to what the sub entries mean.
to find the entry just run
regedit
and search for memory management
Steve Shaw
kru...@acsu.buffalo.edu wrote in article <33B830...@acsu.buffalo.edu>.
..
>> >cache with the *hard drive pagefile* being the actual system
>> >"memory".
>
>> Interesting analogy, painfully true! And it's not a feature -- NT
should
>> make better use of abundantly available RAM.
>
>Well, it really isn't an "analogy"... it's how NT works. If you think
>about it, you will see that as you add more physical memory, your
>pagefile will actually get larger because the LRU (Least Recently
>Used) algorithm will be swapping out more stale pages.
>
>This method is used in almost all "real" operating systems (and
>it works well).
I do see and understand your point. I still say leaving 40% of available
RAM unused, as NT Workstation does, is not good tuning. It slows down
performance, a crucial aspect of a good O.S.
>I can see how a person could say "Hey I have 128 megs of RAM, what
>the heck do I need a pagefile for?". Well, imagine that you could
>install 64 or 128 megabytes of CACHE in your computer... why then
>would you need DRAM? The answer? COST. Megabytes of static ram (fast
>cache memory) costs a LOT more than DRAM. Likewise, megabytes of
>hard drive space are cheaper than DRAM. So, why not let your hard
>drive be your "system memory" and let the DRAM cache it for you?
>
>It really does make sense once you think about it this way...
Well I do like the idea of 128 meg static RAM. Heck, I like the idea of
128 meg, if it's really needed and used. But the question remains: Why
leave half (OK, 40%) of it unused, like NT Workstation does?
Steve Shaw
T. Immel
"Steve Shaw" <nospam_...@kc-primary.net> wrote:
>>I always see you complain about it, but that is what makes NT a
>>true multitasking system. If MS did what you suggest, NT would be just
>>like 95. If you don't like 95, why do you want NT to become 95?
>Actually, one of the good things about Win95 is that, unlike NT, Win 95
>actually uses all the nice juicy RAM in a way that helps performance
NT reminds me of VMS a hell of a lot.
But I *like* VMS.
If you're still reading, Win95 uses RAM nicely
and gosh it's fast and dontcha love rebooting
your Win95 box three times a day when your apps
start crashing into each other?
If I gotta get 96 Meg of Ram to keep Word from swapping
like nuts, so be it! Anything to keep Win95 out of my life.
>Still, NT remains my O.S. of choice. Actually, it's my second choice. My
>first choice would be an NT lets me use all the machine I paid for.
Feel free to call me a windoze N(ice)T(ry) robot. Life's to short
to figure out Linux. (Which also must suck).
Thomas I.
Dave Thomas
Steve Shaw wrote:
>
> Dave Thomas wrote in article <33B7ED...@ml.com>...
> >Steve Shaw wrote:
> >>
> >> kru...@acsu.buffalo.edu wrote in article <33B7EB...@acsu.buffalo.ed
> u>.
> >> ..
> >> >An easy to think about the Windows NT pagefile is like this:
> >> >
> >> >Think of a Windows NT system as having L1 cache, L2 cache and "system
> ram" as
> "memory".
> >> >----------------------------
> >>
> >> Interesting analogy, painfully true! And it's not a feature -- NT
> should
> >> make better use of abundantly available RAM.
> >>
>
>
> >Steve,
> >It is in fact a feature. EVERY multitasking OS does exactly the same
> >thing.
>
> Simply not true. High-performance multi-tasking O.S.'s make use of all
> available memory instead of letting half of it sit idle. I'm not arguing
> which OS is better, or that NT isn't a good O.S. I say that NT could be a
> better, faster O.S. if it would use all it's memory, instead of half its
> memory. (I'm talking NT Workstation here.)
>
> >true multitasking system. If MS did what you suggest, NT would be just
> >like 95. If you don't like 95, why do you want NT to become 95?
>
> Actually, one of the good things about Win95 is that, unlike NT, Win 95
> actually uses all the nice juicy RAM in a way that helps performance
> first choice would be an NT lets me use all the machine I paid for.
>
Steve,
If an OS did as you suggest, that is leave tasks swapped in as much as
possbile to drive maximum use of RAM, every time a new process wished to
start the OS would first have to go out and rim the working sets of
every other task running to free up enough ram to start this task.
Since a OS has no way of knowing in advance when a new task is going to
start, if the OS is designed to be a true multitasking environment, it
will trim as often as possible. If it did not, this leads to a
condition known as "thrashing" wherein every time a page is paged in,
another page must be paged out to make room for it. But then the page
that was just stolen needs to come back in, which forces another page to
be stolen, etc. The entire system grinds to a halt. This is caused by
allowing processes to start without enough virtual storage to support
them. Modern OS'es prevent this condition by not allowing a process to
start unless there is sufficient free ram available, and enough free
space on the page devices. I recently watched a mainframe system with 2
Gigabytes of ram fail to start new address spaces because all virtual
storage was exausted. No matter how much ram you have, it can be used
up. That is why NT and all other multi-tasking OS'es trim pages as
frequently as possible and maximize free ram for use by new address
spaces. What you are describing as ideal is a SINGLE-TASKING OS like
DOS, which such concerns don't enter the picture. Is a single tasking
OS it perfectly permissable to allow all the ram to be used by the
running process because you know another task is not going to suddenly
start. Think about it. If your NT system ust refused to let you start
another window because all your ram was being used by some other program
you were running, you would complain about that too. Multi-tasking
OS'es are different, and people have to get used to them and think of
them differently. This is not DOS. But that is also what gives NT it's
strength. And changing the fundamental nature of an OS from
multitasking to singe tasking goes way beyond the definiton of
"tuning". NT is indeed a tunable system. The resource kit gives plenty
of ways to tune NT. It just can't be turned from a multitasker to a
singletasker which is what you want.
Dave
Steve Shaw
Hi Dave,
You make some interesting points. Since NT may someday conqer the world,
perhaps the issue of NT's performance is worth pursuing....
Dave Thomas wrote in article <33B93C...@ml.com>...
>-----------------
Thrashing is an issue that good multi-tasking O.S.'s have learned to deal
with years ago, some more successfully than others. Yes, all systems have
limitations. However, well designed multitasking systems such as MVS, VM,
Unix certainly do allow full use of RAM, backed by virtual memory. This is
necessary to support a large multitasking workload with good performance.
Paging, properly implemented, is a good thing, but unneeded, premature
paging of the sort that NTW routinely does has no place in a high
performance system.
These facts make it hard to sustain an argument that half the RAM should be
reserved as unused "in case things get busy," as NTW does. Especially in a
multitasking environment, full efficient use of RAM is neccessary for good
performance. A good, fast O.S. can deal with it. NTW can't, yet.
I say to you, to the geniuses at Microsoft, and to my great Aunt Matilda in
Ohio who probably thinks RAM is a new-age rock group, that NTW needs some
work in the memory management department.
And if Microsoft doesn't want to tune NT, or design NT to intelligently
tune itself, then they should at least provide a means to let customers
tune NT to run in a way that best fits each unique workload. Those knobs do
not exist in NT, and they are not in the Resource Kit. NT, and NT's users,
deserve better.
Steve Shaw
dinars
In article <33B93C...@ml.com>, david_...@ml.com says...
>If an OS did as you suggest, that is leave tasks swapped in as much as
>possbile to drive maximum use of RAM, every time a new process wished to
>start the OS would first have to go out and rim the working sets of
>every other task running to free up enough ram to start this task.
>Since a OS has no way of knowing in advance when a new task is going to
>start, if the OS is designed to be a true multitasking environment, it
>will trim as often as possible. If it did not, this leads to a
>condition known as "thrashing"
Not true, the job of any multitasking OS is to run existing
processes as fast as possible, not to swap them out, just
to make short boost on startup of next processs. The memory
reserverd for next process, is wasted memory anyway. Even if
OS reserves some amount of memory to boost next process startup,
imediately after next process is started OS must do the same
job for future startups.
>That is why NT and all other multi-tasking OS'es trim pages as
>frequently as possible and maximize free ram for use by new address
>spaces.
Other multitasking OS'es use LRU to swap out pages, this ensures
minimal harm on already runnings processes. NT uses FIFO, and
any attempt to trim pages as frequently as possible will
cause circular queue of pages swapped in/out.
> NT is indeed a tunable system. The resource kit gives plenty
>of ways to tune NT. It just can't be turned from a multitasker to a
>singletasker which is what you want.
>Dave
NT is only partialy tunable system, for examle no such
tool exists that allows to limit maximum cache size.
Numbers shown in task manager is wrong. My system shows
file cache from 10MB to 15MB, minimal cache size is set
in registry, but when i copy 30MB file to NUL second time,
there is no disk activity at all. Entire file is stored in
RAM, all code/data is swapped out.
Dinars.
dinars
In article <33B830...@acsu.buffalo.edu>, kru...@acsu.buffalo.edu says...
>Well, it really isn't an "analogy"... it's how NT works. If you think
>about it, you will see that as you add more physical memory, your
>pagefile will actually get larger because the LRU (Least Recently
>Used) algorithm will be swapping out more stale pages.
>This method is used in almost all "real" operating systems (and
>it works well).
It works well for almoust "real" operating systems and will work
for NT too, maybe in future versions. Current versions of NT
use FIFO for swaping.
Dinars
dinars
In article <5p86ha$1...@nr1.calgary.istar.net>, ls0...@netcom.ca says...
>A pagefile is used because your system needs more memory
>than it physically has, therefore it swaps out some memory
>temporarily to disk.
Not true for NT. NT uses FIFO for swaping pages out.
Pages are swapped out, because they are "expired".
The general reason for swapping out
(at least on >=80MB system) is because memory pages
owned by cache is newer, not because applications and code
requires more RAM.
>When you create a 40 MB ramdisk, you're decreasing the amount
>of memory available to all your programs, and you'll be
>increasing the amount of swapping, and kill performance.
I use configuration 20MB RAM-swap on 80MB system for long
time, and believe me performance is better.
>your pagefile. Increase your physical memory enough, and you'll
>experience so little paging that the bottleneck in performance
>will be somewhere else (probably you're I/O system).
Any increase of physical memory results in increase of
file cache. My task manager shows that file cache is ~10MB,
but when I copy 30MB file to NUL twice, there is no disk
activity at second try.
Dinars.
kru...@acsu.buffalo.edu
> It works well for almoust "real" operating systems and will work
> for NT too, maybe in future versions. Current versions of NT
> use FIFO for swaping.
Well, "First In, First Out" really means "Least Recently Used". That
is, the first block to be paged to disk is the first one that was used
(and therefore, the oldest).
Roger
--
#################################################
# Roger A. Krupski <kru...@acsu.buffalo.edu> #
# State University of New York at Buffalo #
# 408 Furnas Hall, North (Amherst) Campus #
# Amherst, New York 14260, U.S.A. #
#################################################
Steve Shaw
kru...@acsu.buffalo.edu wrote in article <33BBC9...@acsu.buffalo.edu>.
..
>> It works well for almoust "real" operating systems and will work
>> for NT too, maybe in future versions. Current versions of NT
>> use FIFO for swaping.
>
>Well, "First In, First Out" really means "Least Recently Used". That
>is, the first block to be paged to disk is the first one that was used
>(and therefore, the oldest).
>-------------------------
Nope, "First In, First Out" just means "First In, First Out."
The first page in can easily become the Most Frequently Used, not the Least
Frequently Used, in which case paging it out would have a significant
negative impact on performance.
This matters because performance matters, and I have heard rumors that NT
is being marketed as a high performance system.
Steve Shaw
Alexander Frink
In article <5pf04c$s...@news.latnet.lv>,
din...@mail.bkc.lv (dinars) writes:
> Not true, the job of any multitasking OS is to run existing
> processes as fast as possible, not to swap them out, just
> to make short boost on startup of next processs. The memory
> reserverd for next process, is wasted memory anyway.
Unfortunately, many people perform a benchmark test like this:
Double-click on an icon (e.g. Winword) and wait until the application
has started. The faster, the better the operating system.
No doubt, NT will win.
So, probably we are all using the NT version trimmed for demonstrations
at exhibitions.
Best regards,
Alex
-------------------------------------------------------------------------
Alexander Frink E-Mail: Alexander.Frink (at) Uni-Mainz.DE
Institut fuer Physik Phone: +49-6131-393391
Johannes-Gutenberg-Universitaet
D-55099 Mainz, Germany (sorry for inconvenience due to UCE prevention)
ls0916
dinars wrote:
>
> In article <5p86ha$1...@nr1.calgary.istar.net>, ls0...@netcom.ca says...
>
> >A pagefile is used because your system needs more memory
> >than it physically has, therefore it swaps out some memory
> >temporarily to disk.
>
> Not true for NT. NT uses FIFO for swaping pages out.
> Pages are swapped out, because they are "expired".
This isn't true. NT's Virtual memory manager uses demand
paging, which means that a page is swapped in only when a
page fault occurs, not after a certain period of time.
> Any increase of physical memory results in increase of
> file cache. My task manager shows that file cache is ~10MB,
> but when I copy 30MB file to NUL twice, there is no disk
> activity at second try.
I think this is a pretty general statement. At what level of
RAM are you talking about? Since the cache manager lets
the VM manager control the size of the file cache, it'll grow
and shrink just like the size of a working set (ie. when you're
copying the 30 MB file, it'll probably allocate more memory to
cache if nothing else need it, and then reclaim it when another
process requires memory).
Osvaldo Pinali Doederlein
I have written an article on the subject,
http://www.netpar.com.br/a2.htm
Steve Shaw wrote in article ...
>
> kru...@acsu.buffalo.edu wrote in article <33BBC9...@acsu.buffalo.edu>
.
>.
Steve Shaw
Hi Osvaldo,
I tried to click on the URL you listed, but got a HTTP/1.0 404 Object Not
Found .
Computer performance is fascinating subject.
Alas, my programs are not compiling tonight, I can't figure out why, and
I'm getting tired of messing with the whole thing.
Hope your day is going better....
Steve Shaw
---------------------
Osvaldo Pinali Doederlein wrote in article ...
>I have written an article on the subject,
>
>http://www.netpar.com.br/a2.htm
> Steve Shaw wrote in article ...
>>
>> kru...@acsu.buffalo.edu wrote in article <33BBC9...@acsu.buffalo.edu
>
>
>>.
Willy
On 3 Jul 1997 01:43:40 GMT, din...@mail.bkc.lv (dinars) wrote:
>Not true, the job of any multitasking OS is to run existing
>processes as fast as possible, not to swap them out, just
>to make short boost on startup of next processs.
You haven't worked with real multitasking systems, namely, mainframes.
The job of a multitasking system is to run multiple task. As soon as
the system cannot do this it is useless as a multi-tasking system.
If you want a lesson in RAM usage you should study mainframes. A
small example is batch COBOL programming. You usually have to
calculate exactly how much memory your program will need. You then
specify it in the JCL that schedules your program. You do this
because the mainframe MUST insure that the program can run. That's
its first priority. If the machine thinks that it can't run the
program because it can't commit enough memory then it waits until it
does have enough memory (or, checks priorities to toss something out,
but that's another story)
Willy
On 30 Jun 1997 11:50:08 GMT, The NTGURU <mmont...@earthlink.net>
wrote:
>NT is very smart about it's memory usage and uses whatever
>is not in current use for a disk cache.
Now THAT'S what I thought NT did with it's unused memory.
Willy
>Contrary to
>"popular" opinion on this thread, NT does a less than average job on memory
>management. It really is one of its worst categories.
How did you determine this?
There's one thing for sure. NT runs MS Office processes much faster
than Win95, so it's definitely the best that Microsoft can do.
Brian Martinicky
Hi,
ls0916 <ls0...@netcom.ca> wrote in article <33BC5D...@netcom.ca>...
> dinars wrote:
> >
> > In article <5p86ha$1...@nr1.calgary.istar.net>, ls0...@netcom.ca says...
> >
> > >A pagefile is used because your system needs more memory
> > >than it physically has, therefore it swaps out some memory
> > >temporarily to disk.
> >
> > Not true for NT. NT uses FIFO for swaping pages out.
> > Pages are swapped out, because they are "expired".
>
> This isn't true. NT's Virtual memory manager uses demand
> paging, which means that a page is swapped in only when a
> page fault occurs, not after a certain period of time.
I think the key word was paging *out*. While the NT VMM does use demand
paging stuff into RAM from disk, an LRU algorithm is used for deciding when
things should go out to disk from RAM. The LRU is a little more complex
than a simple FIFO as well.
>< snip >
--
Brian Martinicky
SIG Beringen, Switzerland
bmart...@sigpac.ch
Steve Shaw
Willy wrote in article <33bc8dd2...@news.asan.com>...
>On Mon, 30 Jun 1997 11:13:44 -0700, "David Britton"
><dbri...@cerner.com> wrote:
>
>>Contrary to
>>"popular" opinion on this thread, NT does a less than average job on
memory
>>management. It really is one of its worst categories.
>
>How did you determine this?
>----------------------------
Look at Taskmanager. On NT Workstation, almost half the memory sits
unused.
Frequently used pages are written to disk, only to be paged in again. This
waste of resources and resulting slow behavior is not something you'll see
in a well-designed high performance Operating System.
>There's one thing for sure. NT runs MS Office processes much faster
>than Win95..........
Office 97 apps are quick on either system, but more quick on NT, true. NT
is at its best when an app is in core, doing disk or sometimes video I/O.
Because of the well-publicized architectural changes in NT 4.0 (e.g.
kernal-mode device drivers), disk and video I/O can occur quickly. Note
that NT still doesn't support Mode 3 or 4 EIDE data transfers. This needs
fixing.
NT suffers by comparison when the user moves between multiple programs,
opening and closing multiple files. NT is a decent multitasker, but
nevertheless NT does have a slow "feel" to it when moving between multiple
apps. NT spends to much time flushing its buffers and paging in from disk,
while letting abundantly available RAM go unused.
>..........so it's definitely the best that Microsoft can do.
That's what worries me. But I still hope for better. (Not bigger, just
faster.) NT has so many nice things built in, it could be a really nice,
fast system. It would be a shame if it never reached its potential.
Steve Shaw
dinars
In article <33BC5D...@netcom.ca>, ls0...@netcom.ca says...
>I think this is a pretty general statement. At what level of
>RAM are you talking about? Since the cache manager lets
>the VM manager control the size of the file cache, it'll grow
>and shrink just like the size of a working set (ie. when you're
>copying the 30 MB file, it'll probably allocate more memory to
>cache if nothing else need it, and then reclaim it when another
>process requires memory).
NT kernel is not designed to check "nothing else need it" state of
memory page. There is no hardware support tu check page access
on MIPS platform, and Microsoft choose not to use different code
on intel. The result is - any page are placed in FIFO queue
and last page in queue is target for placing in pagefile. That
active process started 5min ago will be swapped out before
idle process, started later.
Dinars
Willy
>>How did you determine this?
>
>Look at Taskmanager. On NT Workstation, almost half the memory sits
>unused.
Are you running processes that would use all your memory? I have 64
MB of memory. I startup Internet Explorer, Free Agent, then (and
these are the kickers) Lotus Domino server and Lotus Notes client. My
memory usage: 62796k. Now, that would seem like damn near all my
memory is being used. Also, my machine is still very snappy. I start
Notes replication with a remote server, download from the web (two
modems) and every program still feels like it's the only thing
running.
>Note
>that NT still doesn't support Mode 3 or 4 EIDE data transfers. This needs
>fixing.
I not sure where you heard this; all I can do is relay my experience:
I created a 350MB file and copied it from one drive to another. Based
on the amount of time required I calculated around 10MB per second
(IBM Deskstar 4 drives). You simply cannot go this fast with PIO Mode
3 so I must be running with PIO Mode 4 (yes, Mode 3 goes to 13.3MB per
sec but that includes the overhead of making the transfer...I'm taking
about raw data.)
> NT suffers by comparison when the user moves between multiple programs,
>opening and closing multiple files. NT is a decent multitasker, but
>nevertheless NT does have a slow "feel" to it when moving between multiple
>apps. NT spends to much time flushing its buffers and paging in from disk,
>while letting abundantly available RAM go unused.
On my machine it certainly doesn't feel like it's slower with multiple
programs. Lotus Notes is notorious for hogging the disk both in space
and usage. But even with a server and workstation banging on the
disk, other programs (starting them, accessing files) still feels like
they're the only thing running. I really like it.
>>..........so it's definitely the best that Microsoft can do.
>
>That's what worries me. But I still hope for better. (Not bigger, just
>faster.) NT has so many nice things built in, it could be a really nice,
>fast system. It would be a shame if it never reached its potential.
"Better" would be nice but I'm very happy just the way things are (I'm
a not-broke don't-fix kind of guy.)
Steve Shaw
dinars wrote in article <5pjpfq$f...@news.latnet.lv>...
>----------------------
Well, sheesh. That would explain a lot!
NT = Needs Tuning. Really. But I suppose as long as people think NT stands
for New Technology, then they'll also believe it's really High
Performance.
Steve Shaw
The NTGURU
I have to disagree with you on this! The unused memory you see in the
task manager is memory that is being used but not not commited for any
use. NT is able to use all this memory for commonly used tasks. If you
are having frequently used pages pushed to your page file then you
probably need more memory for the type of application you are trying to
use. I have over 100meg of ram in my PC and almost never hit the
pagefile...
Steve Shaw wrote:
> Willy wrote in article <33bc8dd2...@news.asan.com>...
> >On Mon, 30 Jun 1997 11:13:44 -0700, "David Britton"
> ><dbri...@cerner.com> wrote:
> >
> >>Contrary to
> >>"popular" opinion on this thread, NT does a less than average job on
>
> memory
> >>management. It really is one of its worst categories.
> >
> >How did you determine this?
> >----------------------------
>
> Look at Taskmanager. On NT Workstation, almost half the memory sits
> unused.
> Frequently used pages are written to disk, only to be paged in again.
> This
> waste of resources and resulting slow behavior is not something you'll
> see
> in a well-designed high performance Operating System.
>
> >There's one thing for sure. NT runs MS Office processes much faster
> >than Win95..........
>
> Office 97 apps are quick on either system, but more quick on NT,
> true. NT
> is at its best when an app is in core, doing disk or sometimes video
> I/O.
> Because of the well-publicized architectural changes in NT 4.0 (e.g.
> kernal-mode device drivers), disk and video I/O can occur quickly.
> Note
> that NT still doesn't support Mode 3 or 4 EIDE data transfers. This
> needs
> fixing.
>
> NT suffers by comparison when the user moves between multiple
> programs,
> opening and closing multiple files. NT is a decent multitasker, but
> nevertheless NT does have a slow "feel" to it when moving between
> multiple
> apps. NT spends to much time flushing its buffers and paging in from
> disk,
> while letting abundantly available RAM go unused.
>
> >..........so it's definitely the best that Microsoft can do.
>
> That's what worries me. But I still hope for better. (Not bigger,
> just
> faster.) NT has so many nice things built in, it could be a really
> nice,
> fast system. It would be a shame if it never reached its potential.
>
> Steve Shaw
Steve Shaw
The NTGURU wrote in article <4DAB57E229D13BEB.29162957774A742A.8BEB6596EA9B
28...@library-proxy.airnews.net>...
>I have to disagree with you on this! The unused memory you see in the
>task manager is memory that is being used but not not commited for any
>use. NT is able to use all this memory for commonly used tasks.
The unused memory is really used?
Well thanks for clarifying that. <g> I feel better now. Maybe in NT 5.0
Microsoft will change the label from "Available" memory to something more
suggestive of high performance... "Turbo memory" maybe, or at least
something more obscure, like "X-RAM." The imagery has a certain appeal.
>If you
>are having frequently used pages pushed to your page file then you
>probably need more memory for the type of application you are trying to
>use. I have over 100meg of ram in my PC and almost never hit the
>pagefile...
Gulp -- 100 m-m-m-meg? Just to run a PC??
I knew this day was coming. I just didn't think it would be today!
Maybe I should just get with the program and install a gig or so of nice
juicy RAM. Then I could toss the harddrive altogether. That should be
enough to handle NT 5.0, or at least last until I can spring for a new
Cray.
Steve Shaw
Steve Shaw
ft...@aurora.alaska.edu (T. Immel) wrote in article <5p9kkm$v...@news.alaska
.edu>...
>
>NT reminds me of VMS a hell of a lot.
>But I *like* VMS.
>
>If you're still reading, Win95 uses RAM nicely
>and gosh it's fast and dontcha love rebooting
>your Win95 box three times a day when your apps
>start crashing into each other?
>
Of course I'm still reading. Yep, Win 95 is faster, quieter, smoother than
NT in some ways, I admit it. Interesting how Win 95 was considered to be
such a big, demanding Operating System when it came out, but a year later
people think of it as a toy O.S. Progress is a wonderful thing, no?
Actually, Win 95 isn't such a bad O.S. It only crashes when I run my own
buggy programs on it. And it reboots quickly. NT reboots so slow I
have time to get a cup of coffee, but maybe that's a feature.
>If I gotta get 96 Meg of Ram to keep Word from swapping
>like nuts, so be it! Anything to keep Win95 out of my life.
>
96 meg to run Microsoft's Word Processor? Jeepers, I can hardly wait for
Office 98.
>>Still, NT remains my O.S. of choice. Actually, it's my second choice.
My
>>first choice would be an NT lets me use all the machine I paid for.
>
>Feel free to call me a windoze N(ice)T(ry) robot. Life's to short
>to figure out Linux. (Which also must suck).
Agreed. Plus, where are the apps?
I'm not saying NT is a bad O.S. I'm saying it could be better (faster).
Steve Shaw
Willy
> Yep, Win 95 is faster, quieter, smoother than
>NT in some ways, I admit it.
I run Lotus software (Notes, Domino, and SmartSuite) and beta software
from various companies on my NT machine; all of which is buggy and
likes to crash (sorry Lotus). The programs crash all the time. NT
doesn't. I have never had such a stable and fast system. Blows the
doors off of Win95 in every way.
I think people that have problems with Windows NT are blaming the
operating system for their cheap hardware.
Steve Shaw
Mr. Spamasauraus wrote in article <33bf03de....@207.14.113.2>...
>"Steve Shaw" <nospam_...@kc-primary.net> wrote:
>>
>>NT = Needs Tuning. Really. But I suppose as long as people think NT
stands
>>for New Technology, then they'll also believe it's really High
>>Performance.
>
>Well, instead of all this whining, why not go ahead and tune it?
I did. See below....
>
>Just go to www.ntinternals.com and download the freeware cache manager
>for NT4. You'll find that you can fart around a lot and still not do
>much better (and you could do a lot worse) than the default tuning you
>hate so much.
Have you considered the possibility that maybe that's because the freeware
cache manager really doesn't do diddly-squat? See below.......
>Besides including directions on writing your own cache
>control software, it'll will give you total control of all NT caching
>variables such as:
>
>CcFirstDelay: Delay before writing back a page after first access
>CcIdleDelay: Idle period to wait before writing back dirty pages
>CcCollisionDelay: Delay by this amount if writeback not possible
>CcTargetCleanDelay: Standard delay before writing back dirty data
>CcDirtyPageThreshold: Maximum number of dirty pages to allow in cache
>CcDirtyPageTarget: Desired number of dirty pages in cache
>CcAvailableMaxDirtyWrite: Maximum number of dirty pages to write back
>at a time
>CcAvailablePagesThreshold: Require at least this many pages available
>MmSystemCacheWsMinimum: Minimum size of the working set of the cache
>MmSystemCacheWsMaximum: Maximum size of the working set of the cache
>MmDoPeriodicAgressiveTrimming: Agressively reduce working sets of all
>tasks
>MmPeriodicAgressiveTrimMinFree: Trim working set of cache if available
>pages greater than this amount
>MmPeriodicAgressiveTrimMaxFree: Trim working set of cache if available
>pages less than this amount
>MmPeriodicAgressiveCacheWsMin: Trim cache to this amount periodically
>MmWorkingSetReductionMaxCacheWs: Trim cache by this amount if larger
>than max size
>MmWorkingSetVolReductionMaxCacheWs: Trim cache by this amount if
>larger than max size during volume reduction
>
The good folks at www.ntinternals.com don't know for sure what some of
these knobs do. The descriptions are their best guess as to what the knobs
do. They are good enough to admit that. I did try this freeware on both
ntfs and fat, with a variety of parameter settings. I did have hopes for
this program, it seemed like just what the Doctor ordered, but it didn't
make any difference whatsoever with either file system.
It is a major nice idea though, definitely needed.
BTW, sorry if my trying to point out areas where NT could use improvement
seemed like whining. Maybe I've been too insistent. It has seemed
difficult to get people to realize that true High Performance systems
involve more than a slogan printed on a box.
NT is a good system. I had thought that with some relatively minor tuning,
NT could be made to run quite a bit faster, and use less RAM too. One
person has pointed out that NT was built on a hardware platform that didn't
provide hardware support for LRU paging algorithms. _If_ that's true, and
I don't know if it is, then it may not be so easy to effect further tuning
improvements in NT; it's performance will probably remain compromised and
it will continue to gobble and waste RAM. (Hope that doesn't sound like
whining.)
Steve Shaw
Craig Kulesa
The NTGURU wrote:
[snip]
> If you
> are having frequently used pages pushed to your page file then you
> probably need more memory for the type of application you are trying to
> use. I have over 100meg of ram in my PC and almost never hit the
> pagefile...
I have heard the last sentence many times in this newsgroup, and it's
rather frightening. 100 MB to "almost never" hit the pagefile? Why is it
that people (in general, not just the above poster) are so accepting of
software that is unnecessarily hoggish?
Why is it that under Redhat Linux I can run an office suite, Netscape,
various networking services, and my usual set of image processing software
without touching the pagefile -- in "only" 32 MB? In fact, I've worked for
hours on a 64 MB Pentium Pro Linux box aggressively processing data and
never visibly hit outside the I/O cache, much less the pagefile.
Absolutely blinding speed! Regardless of what you think about Linux, it
demonstrates just how fast and efficient a solid, multiuser, multitasking
OS can be.
I've used NT since the pre-3.1 days, and find it a capable OS. But by no
stretch of the imagination is it efficient with hardware resources. It
doesn't HAVE to be this way. Meekly accepting the current VMM screwiness
isn't going to change things. Demanding that MS fix it, or at least
provide (or document) the means of changing its behavior, will.
Craig Kulesa
cku...@as.arizona.edu
http://loke.as.arizona.edu/~ckulesa/
=========================================================================
Steve Shaw
wi...@nospam.com (Willy) wrote in article <33bf0e9...@news.asan.com>..
.
>On Sat, 5 Jul 1997 05:05:18 -0500, "Steve Shaw"
><nospam_...@kc-primary.net> wrote:
>
>> Yep, Win 95 is faster, quieter, smoother than
>>NT in some ways, I admit it.
>
>I run Lotus software (Notes, Domino, and SmartSuite) and beta software
>from various companies on my NT machine; all of which is buggy and
>likes to crash (sorry Lotus). The programs crash all the time. NT
>doesn't. I have never had such a stable and fast system. Blows the
>doors off of Win95 in every way.
NT's stability is one of its strong points, true.
>I think people that have problems with Windows NT are blaming the
>operating system for their cheap hardware.
In many cases, true.
Steve S.
Joseph Malloy
Craig Kulesa <cku...@loke.as.arizona.edu> wrote in article
<slrn5ruhsg....@loke.as.arizona.edu>...
> The NTGURU wrote:
> [snip]
> > If you
> > are having frequently used pages pushed to your page file then you
> > probably need more memory for the type of application you are trying to
> > use. I have over 100meg of ram in my PC and almost never hit the
> > pagefile...
>
> I have heard the last sentence many times in this newsgroup, and it's
> rather frightening. 100 MB to "almost never" hit the pagefile? Why is it
> that people (in general, not just the above poster) are so accepting of
> software that is unnecessarily hoggish?
>
Aye, there's the rub: I gather (having never run it myself (although it
sits on my bookshelf tempting me)) that Linux isn't really ready for the
cost-conscious (support!) consumer and, indeed, many never be (hey, sorry
Charlie, but I'm not ready for either the high prices that Unix
applications tend to cost nor am I ready to "roll" my own (recompiling
freeware)). After my experience with Digital's version of Unix, I'm aware
that it (i.e., Unix, whether Linux or some other flavor) is a demanding
interface -- and I'm just not ready nor am likely ever to be ready to meet
those demands. Windows, whether 95 or NT, allows me to do what I want and
need to do -- and if I don't like Windows, I can always get a Macintosh.
I've got NTW4 and 64 MB (here and at the office) and, as the sentence you
were commenting upon has it, "almost never hit the pagefile" -- and when I
do, it's not glacially slow. It's a matter of what you normally run and
the amount of RAM you have and your tolerance for the relatively fast or
slow speed of disk-subsystems, all of which to say is that it's a very
variable thing depending on the person. Heck, I used to run NTW (3.5x and
earlier!) with a measly 16 MB of RAM (it tended to "hit the ole' pagefile"
a bit more often then!). At home, I've got an ancient (what, 2 and a half
years old?) P-75; at the office, a PP-200. With memory being so cheap and
all, it's heaven (well, as close to it as I'm gonna get right now).
I'm sure Linux is nice, but it's non-trivial and I'm not about to give up
my job just to learn an OS!
- Joe
Steve Shaw
Craig Kulesa wrote in article ...
>The NTGURU wrote:
>[snip]
>> If you
>> are having frequently used pages pushed to your page file then you
>> probably need more memory for the type of application you are trying to
>> use. I have over 100meg of ram in my PC and almost never hit the
>> pagefile...
>
>I have heard the last sentence many times in this newsgroup, and it's
>rather frightening. 100 MB to "almost never" hit the pagefile? Why is it
>that people (in general, not just the above poster) are so accepting of
>software that is unnecessarily hoggish?
>
>Why is it that under Redhat Linux I can run an office suite, Netscape,
>various networking services, and my usual set of image processing
software
>without touching the pagefile -- in "only" 32 MB? In fact, I've worked
for
>hours on a 64 MB Pentium Pro Linux box aggressively processing data and
>never visibly hit outside the I/O cache, much less the pagefile.
>Absolutely blinding speed! Regardless of what you think about Linux, it
>demonstrates just how fast and efficient a solid, multiuser, multitasking
>OS can be.
>
>I've used NT since the pre-3.1 days, and find it a capable OS. But by no
>stretch of the imagination is it efficient with hardware resources. It
>doesn't HAVE to be this way. Meekly accepting the current VMM screwiness
>isn't going to change things. Demanding that MS fix it, or at least
>provide (or document) the means of changing its behavior, will.
>-------------------------
Absolutely true. People who have worked with more than one O.S. know this.
Some uninformed people seem to think that because NT requires massive
amounts of RAM, it must be a "powerful Operating System." This is a
triumph of marketing over engineering reality. The truth is, NT's paging
strategy (local, FIFO replacement) lets RAM sit idle, and is simply
ill-suited for high-performance. NT simplistically tries to limit RAM usage
of potientally bad apps, but penalizes every good app in the process. This
is documented in Helen Custer's book, "Inside Windows NT," page 193.
Steve Shaw.
Stephan Schaem
Mr. Spamasauraus wrote in article <33bf03de....@207.14.113.2>...
>"Steve Shaw" <nospam_...@kc-primary.net> wrote:
>>
>>NT = Needs Tuning. Really. But I suppose as long as people think NT
stands
>>for New Technology, then they'll also believe it's really High
>>Performance.
>
>Well, instead of all this whining, why not go ahead and tune it?
>
>Just go to www.ntinternals.com and download the freeware cache manager
>for NT4. You'll find that you can fart around a lot and still not do
>much better (and you could do a lot worse) than the default tuning you
>hate so much. Besides including directions on writing your own cache
>control software, it'll will give you total control of all NT caching
>variables such as:
Sound great. I wish the site was not "down indefinitely" :(
But then I heard this program is just a 'pretty face' and does nothing.
I really hate when I open 10 time the same directory with a douzen file in
it
that the HD always come on, and that douzen over douzen of page fault
happen.
An OS that require 8 page fault to close emty drawer is very much in need
of tunning.
Stephan
Stephan Schaem
Steve Shaw wrote in article ...
>
> ft...@aurora.alaska.edu (T. Immel) wrote in article <5p9kkm$v...@news.alask
a
>edu>...
>>
>>NT reminds me of VMS a hell of a lot.
>>But I *like* VMS.
>>
>>If you're still reading, Win95 uses RAM nicely
>>and gosh it's fast and dontcha love rebooting
>>your Win95 box three times a day when your apps
>>start crashing into each other?
>>
>
>Of course I'm still reading. Yep, Win 95 is faster, quieter, smoother
than
>NT in some ways, I admit it. Interesting how Win 95 was considered to be
>such a big, demanding Operating System when it came out, but a year later
>people think of it as a toy O.S. Progress is a wonderful thing, no?
People are calling win95 a resource pig... Because it uses huge gob
of resource , it does not make it anything else then a toy to play
directx games
better then NT :)
>Actually, Win 95 isn't such a bad O.S. It only crashes when I run my own
>buggy programs on it. And it reboots quickly. NT reboots so slow I
>have time to get a cup of coffee, but maybe that's a feature.
>
It crashes more often then it should... And you really think 95 boot
quickly?
Some people boot their OS in under 5 second, now thats considered quick.
On my system NT boot basicly the same as 95... on some PPro system NT
take AGES to boot... blue screen forever iddling it seem..
>>If I gotta get 96 Meg of Ram to keep Word from swapping
>>like nuts, so be it! Anything to keep Win95 out of my life.
>>
>
>96 meg to run Microsoft's Word Processor? Jeepers, I can hardly wait for
>Office 98.
>
This is very good news for people using non windows PC! :)
This will drive the price of resource down more. Thanks Bill. (And
every windows user that will upgrade to 128mg of ram at compter r us)
>
>>>Still, NT remains my O.S. of choice. Actually, it's my second choice.
>My
>>>first choice would be an NT lets me use all the machine I paid for.
>>
>>Feel free to call me a windoze N(ice)T(ry) robot. Life's to short
>>to figure out Linux. (Which also must suck).
>
>Agreed. Plus, where are the apps?
> I'm not saying NT is a bad O.S. I'm saying it could be better (faster).
>
Much faster... I almost feel like NT and windows where never designed
to run more then 1 aplication at a time. Even so it can, it does this
not with grace.
Stephan
Bryce Howard Cogswell
In article <uD#Xlmci8...@uppssnewspub05.moswest.msn.net> "Steve Shaw" <nospam_...@kc-primary.net> writes:
> Mr. Spamasauraus wrote in article <33bf03de....@207.14.113.2>...
>
> >Well, instead of all this whining, why not go ahead and tune it?
>
> I did. See below....
>
> >
> >Just go to www.ntinternals.com and download the freeware cache manager
> >for NT4. You'll find that you can fart around a lot and still not do
> >much better (and you could do a lot worse) than the default tuning you
> >hate so much.
>
> Have you considered the possibility that maybe that's because the freeware
> cache manager really doesn't do diddly-squat? See below.......
>
> The good folks at www.ntinternals.com don't know for sure what some of
> these knobs do. The descriptions are their best guess as to what the knobs
> do. They are good enough to admit that. I did try this freeware on both
> ntfs and fat, with a variety of parameter settings. I did have hopes for
> this program, it seemed like just what the Doctor ordered, but it didn't
> make any difference whatsoever with either file system.
>
> It is a major nice idea though, definitely needed.
As author of said program, I can assure you that the program does diddly-
squat, in that it really does tweak the variables it claims to tweak.
The program exports all the variables that we could determine the cache
manager uses. The problem is that the cache manager rarely makes use
of these variables, because the algorithm is hardcoded to only reference
them under specific circumstances which rarely occur. For example, you
can set a maximum working set size just fine, but the code that looks
at and uses the working set variable only gets invoked after the system
is already in a destructive state, and only uses it briefly, not long
enough to get out of the destructive state. This is a property of the
algorithm that cannot be modified without rewriting the code.
Further, many of the variables are only used by the cache manager if the
Aggressive Trimming flag is set, so that may also give the impression that
the program is not working.
You can make up your own mind whether to blame these problems on CacheMan
(the cache tuning program) or the NT kernel.
-- Bryce Cogswell
Steve Shaw
Bryce Howard Cogswell wrote in article ...
>As author of said program, I can assure you that the program does diddly-
>squat, in that it really does tweak the variables it claims to tweak.
>The program exports all the variables that we could determine the cache
>manager uses. The problem is that the cache manager rarely makes use
>of these variables, because the algorithm is hardcoded to only reference
>them under specific circumstances which rarely occur. For example, you
>can set a maximum working set size just fine, but the code that looks
>at and uses the working set variable only gets invoked after the system
>is already in a destructive state, and only uses it briefly, not long
>enough to get out of the destructive state. This is a property of the
>algorithm that cannot be modified without rewriting the code.
>
>Further, many of the variables are only used by the cache manager if the
>Aggressive Trimming flag is set, so that may also give the impression
that
>the program is not working.
>
>You can make up your own mind whether to blame these problems on CacheMan
>(the cache tuning program) or the NT kernel.
>
>----------------
Bryce,
You can be sure the effort and expertise invested in developing that
program is appreciated. I hope to develop performance-oriented system
software myself someday.
My hat is off to you.
Best wishes for success,
Steve Shaw
Michael Seamans
I sit and look at what I started by asking a mere
question... "has anybody tried this before" and look at the
how big this thread has gotten. I didn't ask for any advise
on whether NT can do this or that, nor was I looking for
some self-proclaimed expert to give me a lecture on how NT
handles the paging file, etc.. All I asked was a simple
question. No one looked at what the original post was
looking for. Some just used the thread to "show off" what
they knew about the topic & never addressed the original
question. It's O.K. to offer an opinion, but stretching
this thread out like this is B.S. As for the people who
said that this wouldn't work, let me offer some preliminary
feedback:
1) The RAMDISK utility works very well & the driver starts
at the BOOT level, so the ramdisk is available when the
paging file needs to be created.
2) Although you would need at least 64MB of memory to start
this, you do get a significant boost in processing, but I
need to do much more testing to confirm my data...
3) Some have said that if you have 128MB of memory to try
this out, then why would you need a paging file?? NT is
going to page no matter what. Yeah, NT needs less and less
of a paging file as you feed it more memory, but there
seems to be a happy medium that could be reached by
balancing out your RAM and paging file usage. This is where
your ramdisk would come in. If you're going to page anyway,
why not page to a virtual disk instead of a hard disk?? I'd
take the access time of 60NS vs. 10MS (a factor of 1660%
increase in performance) any day of the week.
As I cited in my original post, this is all in theory & I'm
looking to see if anyone has tried this before...
--
Michael Seamans
SEA...@NOSPAM.EROLS.COM
If replying, drop the NOSPAM from my address
"Some people see things as they are & ask 'Why'; I dream of
things yet to be & say 'Why not'" - George Bernard Shaw
------------------------------------------------------------
------------------------
Michael Seamans <sea...@nospam.erols.com> wrote in article
<01bc84ef$b049d490$1d47accf@developer>...
> I was playing around with the ramdisk utility I found for
> NT (pretty nice...the driver is only 25K in size & can be
> set up in less than 5 min.) and had this idea. Could you
> run NT with all of your PAGEFILE.SYS file on a ramdisk??
> Currently, I am experimenting with ramdisk & I'm going to
> try this theory out. In theory, you should be able to
> create a ramdisk that could be large enough to hold your
> pagefile & every time you boot, your ramdisk is created
> (the driver boots at the BOOT level in NT, so it is
> available when all of your SYSTEM & above drivers start),
> NT creates your pagefile on your ramdisk & it is used for
> all paging. Now to do this you would probably need at
least
> 64-80M of system memory (40M for NT and 24-40M for
> pagefile), but the performance would be outstanding (in
> theory). My question is this - has anybody out there
tried
> this??? If so, what were your results??
>
> --
> Michael Seamans
> SEA...@NOSPAM.EROLS.COM
>
> If replying, drop the NOSPAM from my address
> "Some people see things as they are & ask 'Why'; I dream
of
> things yet to be & say 'Why not'" - George Bernard Shaw
>
>
141.192.83.61
In <5p9kkm$v...@news.alaska.edu>, ft...@aurora.alaska.edu (T. Immel) writes:
>If I gotta get 96 Meg of Ram to keep Word from swapping
>like nuts, so be it! Anything to keep Win95 out of my life.
You are absolutely right, but there is a fix. Win95 _can_ be tuned.
Dynamic cache in Win95 is over aggressive - there is not much sense
to swap apps out to get bigger cache!! Except that you get impressive
I/O benchmark results this way.
And now the cure: in Win95 there is an entry to limit maximum
cache size. Set this to something sensible and your
system will be more usable.
NT would really need this entry, too. Currently there is only
a registry entry to choose between 'small' and 'big' file cache.
The bad thing is that the 'small' file cache is still too big.
Setting 'big' is meant for file server use and little use in
a workstation.
Kari K.
Steve Shaw
Michael Seamans wrote in article <01bc8b4d$31c2bd80$3f6eaccf@developer>...
> I sit and look at what I started by asking a mere
>question... "has anybody tried this before" and look at the
>how big this thread has gotten. I didn't ask for any advise
>on whether NT can do this or that, nor was I looking for
>some self-proclaimed expert to give me a lecture on how NT
>handles the paging file, etc.. All I asked was a simple
>question. No one looked at what the original post was
>looking for. Some just used the thread to "show off" what
>they knew about the topic & never addressed the original
>question. It's O.K. to offer an opinion, but stretching
>this thread out like this is B.S.
Agreed, show-offs are a bore. Sometimes it's hard to tell why people act
as they do, but perhaps someone hoped that by pointing out how NT could be
improved, NT might become a better O.S. for everyone to use. If the day
comes that everyone uses NT, then good, efficient performance might be an
important issue for those poor souls who must live with less than 128 meg
of RAM. After all, which is better, to apply a Ramdisk band-aid, or to
eliminate the (paging) problem that makes the band-aid necessary?
>As for the people who said that this wouldn't work, let me offer some
preliminary
>feedback:
>
>1) The RAMDISK utility works very well & the driver starts
>at the BOOT level, so the ramdisk is available when the
>paging file needs to be created.
>
Great. Glad it worked for you. The Ramdisk caused Microsoft's C++ compiler
to fail on my system, but maybe I'll try it again.
>2) Although you would need at least 64MB of memory to start
>this, you do get a significant boost in processing, but I
>need to do much more testing to confirm my data...
>
>3) Some have said that if you have 128MB of memory to try
>this out, then why would you need a paging file?? NT is
>going to page no matter what. Yeah, NT needs less and less
>of a paging file as you feed it more memory, but there
>seems to be a happy medium that could be reached by
>balancing out your RAM and paging file usage. This is where
>your ramdisk would come in. If you're going to page anyway,
>why not page to a virtual disk instead of a hard disk?? I'd
>take the access time of 60NS vs. 10MS (a factor of 1660%
>increase in performance) any day of the week.
>As I cited in my original post, this is all in theory & I'm
>looking to see if anyone has tried this before...
>
Great. Best wishes for success,
Steve Shaw