NTFS x MFT x CLUSTER

Status
Not open for further replies.

aze

Posts: 37   +0
Hello!
(plz, don't note my bad writing, I'm Brazilian - I don't speak English)
I'm looking for some information.
I have a 40gb quantum Plus AS 7200 [ATA 66] running win2ksp2 (256 ram).
All professional defragmentation programs I tryed (~4) dont defrag NTFS partitions cluster greater than 4kb [mine's 16Kb]

------ NTFS > 4kb cluster ---------
good - fast speed
good - less fragmentation
bad - don't defrag
bad - lost some disk space
------ NTFS <= 4kb cluster ---------
good - can defrag
good - don't lost disk space (relative)
bad - lower speed (relative)
bad - more fragmentation.

============================
C : FAT32 : 32KB CLUSTER : 2GB
D : FAT32 : 32KB CLUSTER : 10GB
E : NTFS : 16KB CLUSTER : 20GB <---
F : FAT32 : 32KB CLUSTER : 4GB
G : FAT16 : 64KB CLUSTER : 2GB
============================

What cluster size should I set for NTFS part? And about MFT? What's the MFT function? What percent of the disk should I set for MFT size (12,5%? 25%?)? Plz, post your comments!
Thanks!
 
Honestly, the difference between NTFS 4kb and NTFS 16kb goes totally unnoticed by the user - But it is nice to know that you have theoretically better disk speeds. ;) Also, the difference shows up in benchmarks.. But really, it is not practical or even useful for use at home.

So my suggestion would be to use the 4kb size since it is more practical and you can defragment.

Have you tried XP/2000's built-in defragmenter to defragment these cluster sizes? It seems silly to me that the >4kb cluster sizes can't be defragmented. Also, I seem to remember Norton Speed Disk 2002 being able to defragment one of my drives that was formatted in NTFS 16kb. You may want to try that program if you haven't already.

A smaller MFT is better, but I didn't realize you can control the MFT size? After a long time, the MFT can get a lot of "junk" entires in it, so I guess it would be possible to get rid of these "junk" entries.. Just remember smaller is better.

Choosing a larger cluster size also makes the MFT smaller, which is the main reason larger cluster size equals better performance.
 
Thanks responding.

After lots of try I know now just the norton speed disk can defrag any ntfs's size cluster 'cause it uses your OWN APIs (not windows's APIs). window's 2k defrager also have the bug with "> 4kb". the solution is norton :( (i prefer Perfect Disk 200)
About MFT you can change it (http://www.pureperformance.com/NT/hd/hd15.htm) easily.
I still not understanding the MFT case. Ones say large is better, others say litte is better. So if I have a 10Gb NTFS part. i'm "obligated to donate" some percent of the disk size to MFT area? how to defragment mft area?

thanks
 
I think you may have a misunderstanding about what exactly you have read.

The link you have posted says it is better to have a larger reserve space for your MFT, but this does not actually change the size of your MFT - Just the amount of space reserved for it -"allocated space' would be a good word for it..

By reserving more % disk space for your MFT, you can keep it from becoming fragmented since it will have more room to grow. If you your MFT outgrows the space you have reserved for it, it may become fragmented and this in bad performance.

This does not change the actual size of the MFT though.. Just how much space is allocated for your MFT.


A smaller MFT is unquestionably better. When your MFT is small, your system has less stuff to sort through and results in better performance. A larger amount of reserved space for your MFT is also probably better. I'm still not entirely clear on the specifics of MFT allocation however.
 
Diskeeper can defragment the MFT and Pagefile when boot time defragmentation is selected. Personally, I'd just use the default cluster size. You'll get more speed with a well defagmented drive than 16k clusters and fragmentation. Disk benchmarks tend to run in an area of disk they can use as a contiguous test block. that's a lot different than having your dlls fragmented all over your hard drive.
 
Status
Not open for further replies.
Back