XP performance deteriorating

[Lucvdv]

I've been (or rather my main development workstation has been) suffering
from a constantly degrading performance for about a year now. It's as if
performace decreases with 5 to 10% each month.

I primarily use Visual Studio 2005 on this machine.

I have an additional "backup" VS2005 setup in a VMWare virtual PC on an
external (USB) harddisk as well, where I keep copies of my sources in case
something goes wrong with the "real" machine.


Both setups are XP SP2, both current with updates, etc..


Now the funny thing is, you would expect that on the same hardware, the
"real" instance of Visual Studio would run (at least) slightly better than
the one in the virtual machine.

Exactly the opposite is true here on my system: VStudio runs almost THREE
TIMES faster in the virtual machine than it does directly on the host.


Starting Visual Studio on the host takes 65 seconds (worst case, it's much
faster when it has already been run and all files are still boing cached).

Starting it in the virtual machine takes 20 to 25 seconds, under the same
conditions (actually, the time of 25 seconds was measured immediately after
a "cold boot" of the VM, and of course the VM's virtual harddisk wasn't
being cached in RAM by the real OS because it hadn't been run yet).


The host has a single core 3.2 GHz processor with hyperthreading and 2GB of
RAM, and nothing else is running when VS is started there that isn't also
running when VS is started in the VM.

The virtual machine is configured with 512M of RAM, and runs on the same
hardware for the rest (except it's physically on an external USB disk,
versus internal SATA in the other case).

 

[smlunatick]

I've been (or rather my main development workstation has been) suffering
from a constantly degrading performance for about a year now.  It's as if
performace decreases with 5 to 10% each month.

I primarily use Visual Studio 2005 on this machine.

I have an additional "backup" VS2005 setup in a VMWare virtual PC on an
external (USB) harddisk as well, where I keep copies of my sources in case
something goes wrong with the "real" machine.

Both setups are XP SP2, both current with updates, etc..

Now the funny thing is, you would expect that on the same hardware, the
"real" instance of Visual Studio would run (at least) slightly better than
the one in the virtual machine.

Exactly the opposite is true here on my system: VStudio runs almost THREE
TIMES faster in the virtual machine than it does directly on the host.

Starting Visual Studio on the host takes 65 seconds (worst case, it's much
faster when it has already been run and all files are still boing cached).

Starting it in the virtual machine takes 20 to 25 seconds, under the same
conditions (actually, the time of 25 seconds was measured immediately after
a "cold boot" of the VM, and of course the VM's virtual harddisk wasn't
being cached in RAM by the real OS because it hadn't been run yet).

The host has a single core 3.2 GHz processor with hyperthreading and 2GB of
RAM, and nothing else is running when VS is started there that isn't also
running when VS is started in the VM.

The virtual machine is configured with 512M of RAM, and runs on the same
hardware for the rest (except it's physically on an external USB disk,
versus internal SATA in the other case).

PCs will slow down if:

1) More programs added to be "autostart" at boot
2) More files stored on hard drive(s) especially the boot drive
3) Anti-virus / anti-malware softwares (with updates.)
4) Fragmentation of hard drives

 

[Lucvdv]

PCs will slow down if:

1) More programs added to be "autostart" at boot
2) More files stored on hard drive(s) especially the boot drive
3) Anti-virus / anti-malware softwares (with updates.)
4) Fragmentation of hard drives

Yes, I completely agree with that.

But in this case the _same_ software on the _same_ PC is running three
times faster in a virtual machine running from an USB drive, than it is
running without anything virtual, directly from the C drive.

Everything that's throwing load onto the system when the virtual machine
isn't running, is also throwing the same load onto the system when the
virtual machine is running.

That should take care of your items 1) and 3), yet Visual Studio starts a
whopping THREE TIMES faster in the virtual machine.

You can omit 4) as well in this case, one of the things I tried to make it
run faster was defragment all drives. It did make a bit of difference, but
less than 10%.


That would leave just # 2), which could explain some difference because
there's hardly any activity there when VS is started in the virtual
machine.

But if having a 80 GB drive slightly more than half full has to explain a
2/3 decrease in speed compared to having it only 20% full, I think there's
something else wrong.


The only thing I'm sure of is that it must be disk related.
Starting VS2005 twice in succession so all dll's etc. are cached in RAM the
second time, it takes more than a minute to start the first time, and only
nine seconds for the second run.

 

[smlunatick]

Yes, I completely agree with that.

But in this case the _same_ software on the _same_ PC is running three
times faster in a virtual machine running from an USB drive, than it is
running without anything virtual, directly from the C drive.

Everything that's throwing load onto the system when the virtual machine
isn't running, is also throwing the same load onto the system when the
virtual machine is running.

That should take care of your items 1) and 3), yet Visual Studio starts a
whopping THREE TIMES faster in the virtual machine.

You can omit 4) as well in this case, one of the things I tried to make it
run faster was defragment all drives.  It did make a bit of difference, but
less than 10%.

That would leave just # 2), which could explain some difference because
there's hardly any activity there when VS is started in the virtual
machine.

But if having a 80 GB drive slightly more than half full has to explain a
2/3 decrease in speed compared to having it only 20% full, I think there's
something else wrong.

The only thing I'm sure of is that it must be disk related.
Starting VS2005 twice in succession so all dll's etc. are cached in RAM the
second time, it takes more than a minute to start the first time, and only
nine seconds for the second run.

Not familiar with VM set ups but don't VM system only work within
RAM??? RAM is always the fastest "media" on any PC.

 

[Ken Blake, MVP]

PCs will slow down if:

1) More programs added to be "autostart" at boot

That depends entirely on *which* program they are. How many is not an
issue. Some of them can hurt performance severely, but others have no
measurable effect on performance at all.

2) More files stored on hard drive(s) especially the boot drive

Not at all true. What's running affects performance, not how much is
stored on the drive.

3) Anti-virus / anti-malware softwares (with updates.)

Depends entirely on the particular software. Although *anything*
that's running has *some* effect on performance, some programs
(including some anti-malware problems) have so slight a performance
hit as to be unnoticeable. Not all security software is as bad as
Norton's.

4) Fragmentation of hard drives

Fragmentation is a factor, but for those using NTFS, it's much less of
a factor than it used to be.

 

[Sloot]

a) I/O operations in and relating to a virtual machine are not confined to
RAM. Programs have to be loaded into memory from a hard drive or other
storage medium before running them, as usual.

The original poster is puzzled because the copy of XP running in VMWare
should be slower. This is because most resources are made available to
programs running in the VM through a virtual layer and this includes the
operating system.

b) RAM is not the fastest storage medium in a PC, registers are.

You, smlunatick, probably shouldn't be posting on this subject.

 

[Lucvdv]

Not familiar with VM set ups but don't VM system only work within
RAM??? RAM is always the fastest "media" on any PC.

No, they run from harddisk.


Anyway, that it's taken so long for me to respond is because I spent a
couple of days backing up, reformatting, reinstalling software and
selectively restoring those parts of the backup I can't miss.

The result is astounding: instead of in more than a minute, VS2005 now
starts in about 5 seconds.


I'm beginning to believe in item 2) from your first reply (total number of
files on boot drive).
Fragmentation etc. were already ruled out, and it can't have been a
misconfigured setting or so because the problem didn't just come into
existence in a single step: it evolved gradually, growing slower and
slower, week after week and month after month.