Upgrading from P4 dual core

JOhnH

Ejection Handle Pulled
Joined
May 20, 2009
Messages
14,214
Location
Florida
Display Name

Display name:
Right Seater
My computer guy built me a server based on Windows 2003 Standard and a dual core Pentium 4 with 3GB memory in 2005. It has an embedded SCZI Raid controller with two mirrored 250GB drives. This system serves a network of 13 XP-Pro workstations and appears to be our source of slowness. Other than a little frustration over performance, t has never caused 1 minute of downtime in over 7 years.

I would like to upgrade the hardware but maintain the OS for now, unless the software is too restrictive on HW. What is the fastest HW config I can put under W 2003 Server?
 
The machine is so far below any of the limitations of Server 2003 that anything newer should be a significant performance increase.

Kinda depends on what you're doing with it. What kind of server is it? Are all files your workstations use saved on it, or is it doing some kind of database or centralized application work?

Keep in mind the Server 2003 and XP are getting long in the tooth, and support for them will slow and eventually end. There's info on Microsoft's website about specific end of service dates.

Best bet is to call up the computer guy and tell him you'd like some upgrade options and explain exactly which things are slow. If the desktop machines are of the same vintage as the server and your applications have been continually upgraded and patched, they're probably much more CPU and RAM hungry than they originally were, and the slowdown may be at the desktops, not the server.

Again, it really depends a lot on what the server is doing.
 
Business computers are not designed to last 7 years. You are on borrowed time. If you are spending the money to upgrade the hardware, you are better off just migrating to Windows Server 2008 R2 now. The labor won't be much different. Just bite the bullet and do it right. How long can your business afford to be down, if the server crashes? How important is your data? Does your email live on the server? Are you adequately backing it up?
 
Actually most servers and systems sold are designed for exactly seven years. It matches the standard IRS amortization schedule. The sales guy will be on your doorstep the day you no longer can depreciate the machine.

Some companies are moving to three year amortization, but not many. Usually the ones in fiscal trouble do it.

I have numerous Linux servers at work that are older than seven years old. (And one hasn't been rebooted in 1227 days as of today.). We're in the middle of a hardware refresh and the new machines are way overkill. We should be doing more virtualization, but there are political and just plain stubborn people who are blocking it in many ways. More boxes, more work for me to do, (which I mostly script anyway) so I just politely disagree and carry on.

Some machines are very capable of running that long. The hard drives, no. Fans, not always. (Ball bearing high quality ones will if kept clean.)

And there are some mid-range enterprise and mainframe folks who'd also disagree too. That stuff gets installed, turned on, and runs for decades. Blade servers too.

In telecom, seeing a 10+ year old Sun server was very common. Suns were built as servers, not as desktop hardware shoved into a rack mount case... but pricing trumps quality and Sun stuff is spendy now and not popular. Especially after Oracle bought them.

It's not always the hardware that needs replacing, it's the never ending merry go round of forced software obsolescence and lack of official support from vendors designed to force you to continually upgrade, which are the limiting factors for businesses.

If you have people willing to maintain older systems and OSs that allow that... running older hardware and just swapping disks regularly is ultra-cheap and can make a significant bottom-line impact.

Microsoft as a general rule, doesn't operate that way.

A side note for the original poster... SCZI... Is actually SCSI. Small Computer Systems Interface. Pronounced "scuzzy" by most folks, your mistake is completely natural. No big deal.
 
Actually most servers and systems sold are designed for exactly seven years. It matches the standard IRS amortization schedule. The sales guy will be on your doorstep the day you no longer can depreciate the machine.

Some companies are moving to three year amortization, but not many. Usually the ones in fiscal trouble do it.

I have numerous Linux servers at work that are older than seven years old. (And one hasn't been rebooted in 1227 days as of today.). We're in the middle of a hardware refresh and the new machines are way overkill. We should be doing more virtualization, but there are political and just plain stubborn people who are blocking it in many ways. More boxes, more work for me to do, (which I mostly script anyway) so I just politely disagree and carry on.

Some machines are very capable of running that long. The hard drives, no. Fans, not always. (Ball bearing high quality ones will if kept clean.)

And there are some mid-range enterprise and mainframe folks who'd also disagree too. That stuff gets installed, turned on, and runs for decades. Blade servers too.

In telecom, seeing a 10+ year old Sun server was very common. Suns were built as servers, not as desktop hardware shoved into a rack mount case... but pricing trumps quality and Sun stuff is spendy now and not popular. Especially after Oracle bought them.

It's not always the hardware that needs replacing, it's the never ending merry go round of forced software obsolescence and lack of official support from vendors designed to force you to continually upgrade, which are the limiting factors for businesses.

If you have people willing to maintain older systems and OSs that allow that... running older hardware and just swapping disks regularly is ultra-cheap and can make a significant bottom-line impact.

Microsoft as a general rule, doesn't operate that way.

A side note for the original poster... SCZI... Is actually SCSI. Small Computer Systems Interface. Pronounced "scuzzy" by most folks, your mistake is completely natural. No big deal.

I didn't say it wouldn't last that long, but they are really not designed or intended to do so. This is the OP's main business server. Seems to me, at the very least he is risking extended down time if it fails (figure at least a full day for recovery to unlike hardware) and the possible loss of data (if he is doing nightly backups and these are complete and successful, the risk is minimized to losing 24 hours worth of data). If backups aren't regular or complete, the risks go way up. In my experience (which is extensive), small businesses rely on their data more than they realize and downtime and data loss can have a significant cost. Now, if you are plumbing company and write up your service orders on old fashioned paper, maybe it is not a concern.
 
I didn't say it wouldn't last that long, but they are really not designed or intended to do so. This is the OP's main business server. Seems to me, at the very least he is risking extended down time if it fails (figure at least a full day for recovery to unlike hardware) and the possible loss of data (if he is doing nightly backups and these are complete and successful, the risk is minimized to losing 24 hours worth of data). If backups aren't regular or complete, the risks go way up. In my experience (which is extensive), small businesses rely on their data more than they realize and downtime and data loss can have a significant cost. Now, if you are plumbing company and write up your service orders on old fashioned paper, maybe it is not a concern.

Understood. That's why I asked what the server is doing. ;)
 
Another issue is using whitebox hardware. If you install on OEM hardware, such as Dell or HP, they will give you gauranteed like parts replacement for the life of the warranty. So, unless you virtualize your server (you probably won't if you are a single server environment), if you had a motherboard failure or such on a whitebox server, you are looking at a possible reinstall of the operating system followed by a restore (depending how your drives are partitioned and applications installed). With gauranteed like hardware replacement, they can swap the part and walk away.
 
Another issue is using whitebox hardware. If you install on OEM hardware, such as Dell or HP, they will give you gauranteed like parts replacement for the life of the warranty. So, unless you virtualize your server (you probably won't if you are a single server environment), if you had a motherboard failure or such on a whitebox server, you are looking at a possible reinstall of the operating system followed by a restore (depending how your drives are partitioned and applications installed). With gauranteed like hardware replacement, they can swap the part and walk away.

I think the mobo would be the only part that would generally require a reinstall if it went belly up and a direct replacement couldn't be found. I've even gotten around that a few times by purchasing replacement boards built around the same chipset.

I suppose a sufficiently obscure RAID controller might also necessitate a reinstall. I don't recall that ever happening, but I suspect it very well could.

Nonetheless, you have a good point. When I used to build custom machines for businesses (which was pretty rarely), I always required that they buy a spare mobo, and I avoided oddball parts in general. I stuck with tried-and-true, mainstream hardware (for example, Intel mobos) for business machines.

I also required that the client buy spare hard drives because the machines were always RAIDed, and I wanted them to have identical spares on hand.

-Rich
 
Back
Top