Chaos

Can anyone explain the advantage or dis-advantage of reducing cluster sizes within XP. Other than the obvious saving of wasted space on files smaller than 4K.

XP allows you to go as small 512 bytes per cluster.

Does the smaller cluster increase performance in any way?

ToshiroOC

They do not inrease performance as far as I have been able to tell, they just save lots of space if you have tons of extremely small files.

__________________
[img][/img]
Peace be with you, Joe.
Driverheaven Staff Member (Supermoderator)

reno

Cluster size is important for random reads and writes, smaller clusters means faster random transfers, bigger clusters means faster sequential reads and writes.

NTFS has MUCH bigger cluster sizes than FAT32 (min in FAT32 is ... 4 BYTES I think? in NTFS it's 512bytes), BUT it fragments less because of this.

And yes, there is virtually no noticable difference if you have a good harddrive. 1024 is what I use for my NTFS partition.

Chaos

Nope I 'm pretty sure Fat32 is 4K not 4bytes thus the smaller sizes with NTFS

sporkpimp

Bigger cluster sizes are better for random reads, but there's barely a difference when you're talking about sequential reads. Which means that, in most cases, you'll get better performance out of bigger cluster sizes, and worse when using smaller cluster sizes (especially when defragmenting for reasons I cannot fully comprehend... I tested min and max cluster sizes on my comp, and it took FOREVER to defrag on 512). Unless your hard drive is <2GB, go for the greedy, space-wasting large clusters... you're not losing that much space at all, and much of that is in your system files... which would benefit from random access speed and easier defragging.

-SporkPimp

Edit: I just had a thought --- does anyone else notice that defragging takes longer on smaller cluster sizes? Or is it just me, or maybe just some random event causing more fragmentation during that install of Windows?

Either way, my advice stands, just that one piece of information is from a single experience of mine.

Veky

I use the default cluster size when formatting. I think it's at 4k now.

reno

I'm perplexed...I always thought FAT32 brought smaller cluster sizes...damn...why does it fragment easier than NTFS if larger clusters cause less fragmentation??

Chaos

Fat32 provided smaller cluster sizes when compared to Fat. Fat16 clusters were 64k, Fat32 took it down to 4K.

sporkpimp

This might be an NTFS-only thing.

For starters, NTFS as a whole works quite differently than FAT does, and the MFT structure causes a lot less drive fragmentation (although I still defrag twice a week, that's mainly because it only takes a half-hour for VoptXP to do all of my drives... even when I've gone months without a defrag, it never hit more than 20% fragmentation on an almost always-on computer).

So, at very least, NTFS generally requires a lot less defragging than FAT (although you do have to defrag the MFT sporadically... no big deal).

As for the cluster sizes... it may not cause more fragmentation, it just might cause a performance hit on defraggers (or maybe even the specific defragger I use, VoptXP)

-SporkPimp

ToshiroOC

What I have always heard is that NTFS is basically FAT32 with some Microsoft mods to make it unusable by anyone else and to make it so that file security was more completely integrated into things... and NTFS doesn't need less defragging in my experience, it simply defrags badly and half-assedly... never use the windows defrag program, try and find an appropriate 3rd party defragging utilitiy, as NTFS won't move all of the files into one contiguous block - it keeps them scattered, which is no way to defragment anything...

__________________
[img][/img]
Peace be with you, Joe.
Driverheaven Staff Member (Supermoderator)

sporkpimp

Even with extremely small clusters, NTFS still defrags faster than FAT ever did, and requires a lot less defragging. I can say that confidently after two years of maintaining four computers in my house, each one running a different OS (mine is XP Pro, dad's is 2k pro, mom's is ME, bro's is a 98lite installation of 98se).

That's the first thing.

Secondly, the security benefits of NTFS are very real. Admittedly, there are easy ways to break it (a Linux boot disk basically defeats every known Windows filesystem security method), but it's a solid addition.

Thirdly, NTFS and FAT32 are related only because they're both from Microsoft. NTFS is actually based off of the filesystem IBM created for OS/2, it works in an entirely different method (notice that there is no MFT on a FAT drive), and it has features that simply can't be implemented in FAT (built-in realtime compression comes to mind... also, the security measures).

*****

I agree with you that 3rd-party defraggers are a must, but not for the same reasons... I use VoptXP, btw, and I would HIGHLY recommend it, since it works rapidly and covers all the functions you'll need, from scanning for errors to optimizing your page file to fixing your MFT.


That said, NTFS partitions do best when files are "loosely packed;" that is to say, defragmented but with small spaces between files.

And, scattering files over the last megabyte or so of drive space has no meaningful performance detriment, since those files are in one place and in the correct order, just with gaps between bytes... gaps between files don't slow down the filesystem like you seem to think they do, what slows it down is when parts of a file are scattered in nonsequential order, forcing the drive to make many more rotations to access the same file.

For example, let's say that 8 byte file ABCDEFGH is in sequential order on bytes 1-8. The disk can then read them all in order. Same thing is there's a space between each byte, because it's still going to hit all of those bytes in a given rotation.

Now, if ABCD is on bytes 5-8, and EFGH is on bytes 1-4, it's going to take at least two rotations to access it.

This is a super-duper simplified situation, of course, but it's a workable analogy.

Super, duper, super-duper simplified.

Like, if x is simplified, and y is a number so big that I can't even type it out on a 32-bit CPU, then this analogy is x^(y^y).

K?

-SporkPimp

ToshiroOC

That makes total sense - you obviously know more than I do about the topic I just have second hand information, and that isn't toally reliable, however, it has been my experience that defragging (at least on my hard drives) takes forever on NTFS - it takes at least 3 days to defrag all of my hard drives with NTFS, while it takes only about a day or so to get through all of my drives (while they were still FAT32)... I have 4 hard drives, 2 IBM 120GXP's and 2 Maxtor Diamondmax...

__________________
[img][/img]
Peace be with you, Joe.
Driverheaven Staff Member (Supermoderator)

Vampyromaniac

A day or so?? Lol, do you defrag just once a year?

__________________

ToshiroOC

No, about once every other month, mainly because it takes so long. I have about 300gb of data, and they are pretty big files along with thousands of little ones...

__________________
[img][/img]
Peace be with you, Joe.
Driverheaven Staff Member (Supermoderator)

Louie6666

Hmmm something wrong there m8! takes me less than 5 mins to defrag 40gigs once a week with NTFS! I think you should set scheduled defrags in middle of nite are you using diskeeper, or builtin defrag?

I always found fat32 slower to defrag! have you got NTFS set to 4k clusters lol

__________________