raid 0 stripe/cluster size question?

imported_cmltech

Junior Member
Jan 7, 2007
1
0
0
hey guys, I have setup 2 500gig WD RE2's in a raid 0 with a 128k stripe size. 128k was the largest that the sil bios would allow. now once getting into windows and going to the disk manager, I've set it up as a primary partition formatted in NTFS and selected the "default cluster size". now windows default cluster size is 4k if i recall. so now does windows see this partition as 128k clusters or 4k clusters.

little background here. this machine is a intel xbx with a core 2 duo 6800 with a 150gig raptor as the boot drive. i do video editing and hd captures on this machine. so the 1gig raid 0 is for obviously hd data capturing/storage for editing. file sizes range from 300meg to 15gig so a large cluster size will benefit me more.

now on the board is a sil 3120 raid controller (i believe) however I also have a promise fasttrak S150 TX4 raid as well but I'm guessing the SIL is more powerful and faster being that it's so much newer? (please if anyone thinks the promise is better than the onboard SIL let me know)

so with 128k cluster sizes being the largest that you can set(on both the SIL and the Promise), setting up the raid 0 in their respecting bios leaves it at 128k? or when you add the disk in the disk manager can you change it to a larger or smaller size?
 

JAG87

Diamond Member
Jan 3, 2006
3,921
3
76
ok, ill try my best to explain

you have 4K NTFS clusters. this means that no matter how big or small a file is, it will be written in 4K clusters on each disk. Therefore if a file is 16K, it will be written using 4 clusters, and if its 1K it will be written in a 4K cluster (taking up the whole 4K of your hard disk)

you have 128K RAID clusters. this is the smallest amount of data that will be written on a single disk. anything bigger than 128K will be striped and split into 128K parts. so if you have a 512K file, it will be written in 4 clusters of 128K, 2 on one drive and 2 on the other. The same problem as with the NTFS clusters occurs, in which if you have a file smaller than 128K, it will be written on one drive and it will take up a whole 128K (when in reality its only taking up 4K, your raid controller will think its taking 128K). Thats why big striping sizes are not good for small files, because of all the slack. But they are very good for large files because they accellerate reading and writing speeds.

so they are really the same concept, but at the end its the RAID clusters that make the difference in slack, since they are almost always bigger than the NTFS culsters.

hope that helped :)