New CU/Mobo, Same System: Any Hope ?

My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

If I were to simply put the existing system on a new mobo/faster CPU and
figure on installing the new drivers for the new mobo (instead of
re-formatting C: and re-building from scratch) what would my chances of
success be?

50-50, I'd give it a shot before doing it the right way.
--
Pete Cresswell

"(PeteCresswell)" wrote

| If I were to simply put the existing system on a new mobo/faster CPU and
| figure on installing the new drivers for the new mobo (instead of
| re-formatting C: and re-building from scratch) what would my chances of
| success be?
|

I think this might have been talked about recently.
On XP I keep a disk image with the IDE drivers removed
and all the software iinstalled and set up. That saves
a lot of work.

If I copy a disk image without removing those
drivers, moving to a new motherboard, it results in
a bluescreen and a lot of work with boot disks. I
don't know whether Win7 requires the same care.
I'd do that to be on the safe side. Remove the drivers.
Put in generic. Shut down without rebooting so that
Windows can't put them back. Then make a disk image
and copy that to the new machine. Or move the disk
to the new machine. (After making a disk image anyway.)

Other drivers seem to be OK. You'll probably have new
audio, graphics, USB and all that. The motherboard CD
should have those drivers. But Windows doesn't really
complain about those. The biggest problem is batting
away all the hardware wizard prompts until you're
ready to deal with them.

(PeteCresswell) wrote:
My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

If I were to simply put the existing system on a new mobo/faster CPU and
figure on installing the new drivers for the new mobo (instead of
re-formatting C: and re-building from scratch) what would my chances of
success be?

50-50, I'd give it a shot before doing it the right way.

If this was Win2K OS (no activation), this would
be a worthwhile topic of discussion.

On OSes with activation, the reaction to the change
can be completely hostile (OS locks up). Or, you
will be given 72 hours to re-activate.

On Win2K, all you had to do, was match the disk driver.
If the original moth3erboard had an ICH5 and you had
used some IDE driver, you would use a newer motherboard
with another Intel Southbridge, and set it in IDE mode.
And that was to try to ensure it would boot.

If that wasn't feasible, you could stick a removable
storage controller into a PCI slot. I used to use
my Promise Ultra133 card for that. Then, move the
card from the old mobo to the new one, leaving the
C: drive connected to the card. It was certain to boot
when it got to the new setup, because its "friend"
had come with it. Later, you could install the rest
of the drivers on the new platform, move the disk cabling
to the Southbridge, disconnect and remove the Promise
card. I used to call that a "bounce" install, as the
Promise card "bounced into the picture, then bounced
back out again" when it was all done.

But later OSes with activation, you might have to "Seal"
the OS. And I don't know all the IT guy incantations
for doing stuff like that. Presumably this takes
some application of WAIK/WADK kit. And I would have
to learn stuff.

*******

I would rather fix the problem on the existing
hardware.

First you ask yourself, whether the hardware is
"mighty enough" for the job. And if that superficial
analysis is half-ways correct, you can then spend some
time figuring out why it is so slow.

For example, say you'd built your server with an Atom
processor running at 1.1GHz and you had 512MB of RAM.
Yes, that would suck. And we'd run off and buy something
a wee bit better. But if you have a Core2 or better,
I'm willing to bet you could get something reasonably
out of it.

If you're doing RAID5 in software, you're asking the
CPU to do XOR calculations for the data. It may be
better to use a real RAID card for that, because
it leaves a bit of CPU left over for other things.
You might only get 100-150MB/sec out of a software
RAID5 say. Consider how "data-full" the task is,
and whether you're asking the CPU to do a lot of work
or not. Basic I/O, like reading a single hard drive and sending
the sectors over a network, should take nearly no CPU
to do, as a lot of it is done with DMA transfers. Only
if the data needs to be re-packed before transmission,
would it need a few percent CPU. However, more fancy
data manipulation (Storage Spaces, storage extenders,
redundancy, shuffling and rebalancing storage arrays),
some of that is more demanding. And you know how
modern OSes are fully capable of amusing themselves,
and wasting CPU cycles with almost no help at all.

Paul

On Sat, 03 Dec 2016 11:44:03 -0500, "(PeteCresswell)"
wrote:

My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

If I were to simply put the existing system on a new mobo/faster CPU and
figure on installing the new drivers for the new mobo (instead of
re-formatting C: and re-building from scratch) what would my chances of
success be?

50-50, I'd give it a shot before doing it the right way.

Back up your drive using cloning software before you do anything else.

Then, just before removing the old mobo, you might want to investigate
making minisetup run on next boot by running SysPrep. Some of the
procedure outlined here will be overkill for you, and also it's based
on Win2k so obviously it's out of date, but the general procedure
should be similar in W7:

http://www.macfh.co.uk/JavaJive/Wind...eBuilding.html
--
================================================== ======
Please always reply to ng as the email in this post's
header does not exist. Or use a contact address at:
http://www.macfh.co.uk/JavaJive/JavaJive.html
http://www.macfh.co.uk/Macfarlane/Macfarlane.html

On Sat, 03 Dec 2016 11:44:03 -0500, "(PeteCresswell)"
wrote:

My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

I'm with Paul. I'd want to fix the issue instead of throwing random
hardware at it.

There are people here who are willing to help.

--

Char Jackson

"Paul" wrote

| If this was Win2K OS (no activation), this would
| be a worthwhile topic of discussion.
|
| On OSes with activation, the reaction to the change
| can be completely hostile (OS locks up). Or, you
| will be given 72 hours to re-activate.
|
You're assuming he doesn't have a full license.
In that case, yes, he's out of luck. But if he does
have a full license he's allowed to put it on as many
machines as he likes, as long as it's one at a time.

(PeteCresswell) wrote:
My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

If I were to simply put the existing system on a new mobo/faster CPU and
figure on installing the new drivers for the new mobo (instead of
re-formatting C: and re-building from scratch) what would my chances of
success be?

50-50, I'd give it a shot before doing it the right way.

IMO, you will be spending countless hours trying to get it to work and then
end up formatting and installing from scratch anyways.
Might as well do it the right way from the beginning.

A core-2 machine should be sufficient as a server.
You did not say how many other machines and communication connections
it serves simultaneously though. 10, 40, 100?

On Sat, 03 Dec 2016 13:16:24 -0600, Char Jackson wrote:
On Sat, 03 Dec 2016 11:44:03 -0500, "(PeteCresswell)"
wrote:

My Intel Core-2 backup server has been maddeningly slow since Day-1.

I'm thinking about a new mobo/faster CPU.

I'm with Paul. I'd want to fix the issue instead of throwing random
hardware at it.

+1

I don't have the knowledge to contribute to fixing this particular
problem, but based on my experience with software I endorse Paul and
Char's approach. If you switch motherboards but keep your existing
Windows, even without considering the activation issue, there's a
significant chance that you'll also be keeping the problem.

Have you checked whether all your existing drivers are up to date?
It's quite common for PC manufacturers to freeze an image and install
it on a long run of manufactured PCs, even after new versions of
drivers have been released.



--
Stan Brown, Oak Road Systems, Tompkins County, New York, USA
http://BrownMath.com/
http://OakRoadSystems.com/
Shikata ga nai...

Per Paul in Houston TX:
IMO, you will be spending countless hours trying to get it to work and then
end up formatting and installing from scratch anyways.
Might as well do it the right way from the beginning.

That rings true to me... case closed... especially since the number of
apps on the machine in question is minimal.

But your "Fix what is there" observation also rings true...... but I've
already re-formatted and re-built once.... to good effect: before it was
absolutely heinous, now it's simply irritatingly slow.


A core-2 machine should be sufficient as a server.
You did not say how many other machines and communication connections
it serves simultaneously though. 10, 40, 100?

I really do not see much CPU usage... and minimal connections: all it
does is mirror my NAS box.

That being said, I am running a sort of faux-RAID application called
DriveBender. Hot-swappable and, supposedly, I can lose two drives
without losing any data.

It's claim to fame is that it does not care if the drives are
different.... I just kept piling them in there until I had enough
terabytes to mirror the NAS box with dual redundancy.

If I were looking for a hardware issue in the existing system, my money
would be on a flaky drive.... Found one of them already some weeks back
and removing it helped.

I guess that, given a flaky drive can slow the whole thing down, with 15
drives my chances are fifteen times greater....
--
Pete Cresswell

(PeteCresswell) wrote:


I guess that, given a flaky drive can slow the whole thing down, with 15
drives my chances are fifteen times greater....

But you should be checking the SMART stats
on each drive *now*, so you will have a
reference point in future to compare to.

Not only do I use SMART, I also benchmark
the drive. And see if there are any downward
spikes in a read-only surface scan. As that
indicates trouble before SMART makes a
note of it. SMART works well if surface
defects are evenly spread over the platter.
If the defects are all in one small area,
that area can cause performance issues, and
the SMART indicators will still be "all green".

(Example of a drive with a bad spot in it. On
the platform this drive is on, it happens to be
"bus-limited", which accounts for the flat shape
to the curve. The classical zoned behavior is seen
from 80% to 100% on the capacity curve.)

http://lh4.ggpht.com/eucaly61/SQCFxS...SU%20MHT2060AH[6].png

Sometimes, a drive will "perk up" if you rewrite
the surface. And backup and restore of all
the sectors, will do the job. (I use "dd"
or "ddrescue" for this.) It's not even clear
to me, what this is doing, as I don't think the
'magnetism' on the drive is bad, and if there
was actual surface damage, there should
still be artifacts later. (No, I'm not a
subscriber to any Spinrite theories either.)

Paul

Per Paul:
But you should be checking the SMART stats
on each drive *now*, so you will have a
reference point in future to compare to.

I will start doing that.

In the past I have been intimidated by SMART because of my general
cluelessness.... but keeping the numbers and looking for changes, I can
handle.... -)

Tangentially:

I just found a drive with 542 weak sectors - which seems to me tb pretty
extreme.

- Can "Weak" sectors cause problems? Looking at it from the outside,
I would think they might.

- Is there a way to force remapping of weak sectors ?


--
Pete Cresswell

(PeteCresswell) wrote:
Per Paul:
But you should be checking the SMART stats
on each drive *now*, so you will have a
reference point in future to compare to.

I will start doing that.

In the past I have been intimidated by SMART because of my general
cluelessness.... but keeping the numbers and looking for changes, I can
handle.... -)

Tangentially:

I just found a drive with 542 weak sectors - which seems to me tb pretty
extreme.

- Can "Weak" sectors cause problems? Looking at it from the outside,
I would think they might.

- Is there a way to force remapping of weak sectors ?



There are two counters of interest.

Current Pending Sectors
Reallocated Sector Count

Current Pending Sectors are sectors that need to
be tested. On the next write, there will be
an opportunity to test them.

Reallocated Sectors is what happens, if the Pending
Sector fails to write properly, and cannot be read
afterwards. A spare sector from the same area of
the disk is used to replace the duff one.

This talk of "Weak" sectors, means some software
may be looking at the Current Pending. Not all
drives handle Current Pending the same way. Most
of the drives here, you never see the Current Pending
go non-zero. The functional description of what
is supposed to happen, implies first you see
Current Pending, and then a bunch of those
get turned into Reallocated. But the experience
here is different, with stuff just ending up in
Reallocated immediately.

So on mine, the Reallocated is the one to watch.

Reallocated is thresholded. Even if the drive
has swapped out 100,000 bad sectors, it still
reads zero. Then, for the last 5000 spares the
drive has got, the Reallocated counter goes non-zero.
The design is intended to prevent people from
"cherry-picking" drives. The counters
are adjusted on purpose at the factory, so
all the drives are "identically good". No
drive actually leaves the factory with zero
defects - I can assure you of that. But the
counters on all the drives show zero when they
leave the factory. (Obviously there is a counter
inside the drive, that shows the true value.)

*******

You can also use the transfer rate curves as
an early warning. These two drives are good.
But I've had one 500GB Seagate, where it only
delivered 10-20MB/sec writes, and the Reallocated
was still zero. I had to replace the drive, because
it was slowing down the OS too much.

http://s29.postimg.org/8b7cj872v/wd500gb.gif

You'll notice in the trace, there are some tiny
spikes. Some of this has to do with interference
from background OS processes. An "important" spike might
be rectangular, 50GB wide, and align with where the
OS is located. Whereas a much thinner spike or smaller
spike, might be a testing artifact.

I have slightly better luck testing on Win2K, where
the OS doesn't play as much Solitaire in the background
when I'm trying to run HDTune :-)

This drive is my star pupil. This drive doesn't spike like
the other kids. You see only the zone recording pattern.
So this is what the curve on your drive is supposed to
look like. As of today, this drive is up around 35000 hours
of life, and is still clean. I have no idea what's
up with this drive! Why is it so good ? What is the
secret to its health ? Win2K OS lives on this drive.

http://s9.postimg.org/qctigqo4f/ST35...6897_hours.gif

Paul

On Sat, 03 Dec 2016 21:33:19 -0500, "(PeteCresswell)"
wrote:

I really do not see much CPU usage...

That's an important piece of information, especially considering that
you started the thread by wondering if you should replace it. Now we
know that you should not replace it, thus already saving you money. ;-)

Seriously, you've said the system is slow, but I don't think you've
described how it's slow. What is it doing that makes you think it's
slow? Which subsystem is slow? We know it's not CPU, and I'm guessing
it's not memory or video, does that just leave networking? If so, we can
rule out a motherboard, your other suggestion, thus saving you even
more. Without any real info, though, it's hard to make much more
progress.

Decide which subsystem seems to be slow, then test that subsystem to see
if it really is slow.

--

Char Jackson

On Sat, 03 Dec 2016 21:33:19 -0500, "(PeteCresswell)"
wrote:

I am running a sort of faux-RAID application called DriveBender.
Hot-swappable and, supposedly, I can lose two drives
without losing any data.

I've been using DriveBender since 2009 and like it a lot, but I would
never call it "Hot-swappable". It complains mightily if a pooled drive
drops out or gets pulled out of the system, to the point where it falls
back to a read-only mode. Further, replacing a failed drive or adding a
new drive doesn't automatically put that drive into the pool. That's a
manual process, and if you have drive leveling enabled, DB performance
will take a hit while DB populates the new drive with files from the
other drives.

As for being able to lose two drives, that's only if you're doing
mirroring on every single one of your folders. You can select your
top-most folder for mirroring and everything below it comes along for
the ride, but I'm not going to assume that you've done that.

--

Char Jackson

Per Char Jackson:
As for being able to lose two drives, that's only if you're doing
mirroring on every single one of your folders. You can select your
top-most folder for mirroring and everything below it comes along for
the ride, but I'm not going to assume that you've done that.

Yes, that is what I did.... and DB's usage chart suggests that all files
are mirrored.

As far as subsystems go, I am in the process of doing an HD Tune
"Benchmark" test for each of the 15 drives. Only have 5 done so far,
but I am pretty sure I have already found one problem drive:
http://tinyurl.com/jltzr3z
https://photos.google.com/album/AF1Q...3DbHoWhO-2-SyX

--
Pete Cresswell