How often do you defrag?

[Phoenix86]

Originally posted by: BikeDude

Originally posted by: Phoenix86
The systems I was supporting a few years back (NT4) had horribly low memory, and thus used the page file quite a bit. Often over time the PF would become very fragmented and the systems would generate errors. Defragging the HDD would stop these errors. Eventually we worked it out that removing the PF, and re-adding it on a clean partition would solve it as well.

One look at driver verifier (verifier.exe) will tell you that a lack of system resources (memory) is one of the main culprits when it comes to crashes in device drivers (-> BSODs). Specially NT4 was plagued by this, which is why Windows 2000 was touted as a much more stable platform in the first place; MS had extended their QA work to third parties and put lots of drivers through vigorous testing.

So sure, defragging the pagefile probably helped, but I very much doubt that the fragmentation was the root cause; I think you guys only accomplished to mask the problem (a faulty device driver).

A colleague of mine had stability problems until he pulled one of his DIMMs thereby disabling dual-channel mode for his P4 (i865 motherboard). He is happy now, but do I think dual-channel is really to blame? No, he probably just masked the real culprit, whatever it was. :/ (I'm not going to conclude anything either way)

Perhaps, but the driver worked 100% otherwise. This was an imaged based setup running on thousands of PCs, so they all had the same drivers (within the hardware set for the different models). The drivers were updated ~6 months, and no image was "immune" to this problem. I'm just throwing that out... You should cover all the bases when TSing, and if the system is unstable, defragging it won't hurt. It's the things you don't think could affect the problem that you'll never test for, and will bite you in the a$$ the worst.

 

[spyordie007]

Windows' default (10% of available hard disk space used as cache or something?) is ridiculous.

AFAICT IE doesn't clean temp files continuously, but once every day or so. In any case, mine would sometimes stall for a good 5-10 seconds, presumably because it was cleaning those files (1GB+ worth). I reduced the cache size to 20MB and the problem went away.

So, to a certain degree, I'll buy into the "clean temp files" idea, but I never (repeat: never) manually delete that folder. There has never been a reason to, because after reducing the size it simply works. (on my three machines)

I agree that it's way too much. On our network I've actually used a GPO to force everyone's cache to stay smaller.

I've heard a couple people now say that MS will be addressing this in IE 7; but we'll have to wait to see if that's really the case.

 

[TGS]

I've always wondered why the 10% of disk space stuck around so long.

Perhaps back in the 3-5GB era certainly, 300-500 MB. Though with people using giant hard drives as single partitions I can see the IE cache taking up gigabytes of space, which makes no sense what so ever.

 

[ZLRider]

Around once every 2 weeks depending on work load and memory... could be longer now that I think about it. :roll:

 

[TartuAngel]

since i try to keep my pc in order all the time i dont defragment the hard drive before i notice any slowdowns. That is maybe 3-4 times a year. A good thing is to get rid of all programs you haven't used for over 2 months The less files you have the better.

 

[ProviaFan]

Originally posted by: TGS
I've always wondered why the 10% of disk space stuck around so long.

Perhaps back in the 3-5GB era certainly, 300-500 MB. Though with people using giant hard drives as single partitions I can see the IE cache taking up gigabytes of space, which makes no sense what so ever.

I wonder if it didn't also have to do with dialup traditions, where having a larger cache might have been beneficial. With an increasing amount of high-speed network connections, a smaller cache would be just as good if not better.

 

[imported_BikeDude]

Originally posted by: Phoenix86
Perhaps, but the driver worked 100% otherwise.

Well, that's true of any driver that causes frequent/infrequent bluescreens...

The correct approach is securing memory dumps, attempt to pinpoint the problem device driver and report your findings to the vendor in question. If that doesn't pan out, then you try the drastic workarounds. You guys started with the latter, overlooked the former and afterwards managed to advocate a rather strange hypothesis. Not to mention that if the fragmented pagefile was really to blame, it would've been news at the time (and something the OS vendor would have had to address).

 

[imported_BikeDude]

Originally posted by: spyordie007
I've heard a couple people now say that MS will be addressing this in IE 7; but we'll have to wait to see if that's really the case.

I just hope they fix the way they download files. A file is first written to the tempfiles folder and then copied to the destination location. Quite a bit slower than simply writing the file to the correct place in the first place.

 

[ProviaFan]

Originally posted by: BikeDude

Originally posted by: spyordie007
I've heard a couple people now say that MS will be addressing this in IE 7; but we'll have to wait to see if that's really the case.

I just hope they fix the way they download files. A file is first written to the tempfiles folder and then copied to the destination location. Quite a bit slower than simply writing the file to the correct place in the first place.

I'm not sure why they do that. It might have made a tiny bit of sense back in the day when only one hard drive was normal, and moving the file involved a quick update to the FAT or MFT. However, it's kind of stupid now to download a huge file and then have to wait further for the system to shuffle it from one drive to another, or from my drive to another drive on the network.

The only reason I could think to do this would be if the file were to be saved to the desktop. In that case, the file would appear as soon as IE started writing, and given the crowd that IE is targeted at, I wouldn't be surprised if someone tried to click on the file before it was finished, causing corruptionarity to ensure. Of course, they could write a special case for if the file were saved to the desktop (bad idea), or just set the hidden attribute until the file is "done."

 

[The Pentium Guy]

Once a month.

 

[powerMarkymark]

When needed.

M@rc

 

[imported_BikeDude]

Originally posted by: ProviaFan
I'm not sure why they do that. It might have made a tiny bit of sense back in the day when only one hard drive was normal, and moving the file involved a quick update to the FAT or MFT.

Moving/renaming a file is still a cheap operation, but I think IE actually copies it... (I usually download to the "wrong" drive, so I'm not 100% sure)

Of course, they could write a special case for if the file were saved to the desktop (bad idea), or just set the hidden attribute until the file is "done."

There are several mechanisms that could be used, from the simple one of naming the file "download_in_progress#1" to simply creating the file without any sharing enabled (hence preventing other processes access to the file until its file handle is closed).

But I guess the real reason is to not bypass the cache. I.e. if you download the same file multiple times, the second time will be faster... (not very useful) Plus, IE can internally treat all files the same way and worry about location later. (aka "the lazy school of software development")

 

[Nothinman]

Moving/renaming a file is still a cheap operation, but I think IE actually copies it... (I usually download to the "wrong" drive, so I'm not 100% sure)

Yes it does and it's extremely stupid. They should either move the file or create a hard link to it, both operations are instantaneous to the user.