Is overlapping two RAID5 arrays on same drives a bad idea ??

is placing two RAID5 arrays on disk as shown below Is advisable? Will this create performance problems?

[SIZE=2]

            sda-(500GB)        sdb-(1TB)       sdc-(1TB)         sdd-(1TB) 
              (250MB)----------(250MB) ---------unused------------unused------->(/dev/md0) RAID1
              (470GB)----------(470GB) ---------(500GB)-----------(500GB)------>(/dev/md1) RAID5
               unused----------(500GB) ---------(500GB)-----------(500GB)------>(/dev/md2) RAID5

It depends. For performance reasons, it is a horrible idea. RAID 5 is expensive for writes, and unless the drives are particularly fast (based on your drive sizes, I would say they are SATA or nearline SAS) the performance penalty will be visible.

Also, you risk losing more data this way. You might be more interested in doing a RAID 1E or 10. You lose more space this way, but it is far more resilient and faster.

you`re right mark they`re SATA drives ,
although i thought that the performance would be an issue here, i needed a spot light on it , but unfortunately i`m compelled to make use of those 4 disks with a RAID-5 implementation to get advantage from space / Write speed , can u imagine a better partiotionnig theme , considering that i`ll be in need for a ,
1) /boot on RAID-1
2) one big resilient LV /backup to hold the backed up Data across the network.
its a Power-Edge-T110 with 4 SATA controllers ,
-4 HDD, 1x500GB , 3x1TB ??
-CentOS 5.6(Final) x86_64 2.6.18-238.9.1.el5xen installed ,holding Amanda as a centred backup Server

thanks again Mark , i do appreciate it ..

Well, unless you are actually planning on using the xen functions, I would personally use the vanilla kernel.

The other options, I feel would be to use the 500GB drives in RAID 5 (chunk size should depend on the size of the majority of your files and the type of file system.

For the 1TB drives, a RAID 1E might make more sense. The issue is not only about the speed of the drives in optimal state, but how quickly you can recover from a failure before another read failure causes data loss. RAID 1/10/1E call for a simple copy, where RAID 5 and 6 usually call for an xor checksum.

With larger and larger drives, and the long time to get spares, a failure can easily go through more than one drive and cause data loss.

You may also want to consider using something like DRBD to another machine, instead of just amanda if your data is critical. Alternately, unison works well when the environment is mixed.

well , i haven`t got the time to try-out a vanilla kernel yet, its almost my 6th month with linux as i`m continuing to learn throughout the community beside my Self-Study , but i feel happy with a redhat based distro as a start + i might make use of the virtualization add on centos .
thnks for mentioning ur own opinion about vanilla .

i know it `d be stupid to ask but would u please give a simple example on how can i calculate that ? i `d like to know from you
also could u give me an example partition theme using Raid1E like the code i posted ?
i know how to build a RAID1, RAID5 Array , but i didnt came across 1E before , if there any how to for this i`ll be interested to read about it .
thanks a lot mark for helpin ..

Larger files, on average, should have a larger chunk size, so that you can satisfy a small request with one drive, allowing alternate requests to be satisfied by other drives. I tested with things like fio, and bonnie++ to get results with the devices I was using and eventually settled on a 256K chunk size for my video collection, which is averaging 1GB per home movie.

For my /boot partition, I went with 8K, because the files are tiny. I tried 64K to 2MB chunk sizes, in various RAID flavors before settling on RAID 1E.

Also, your partition layout does not have to matter much when you are using RAID, albeit with a few exceptions. Your /boot volume cannot reside on a RAID 5 or 6. I believe only mirroring is currently supported.

Also, when I said a "Vanilla Kernel" I was not referring to a particular distribution called Vanilla, but rather, Vanilla has a connotation of being stock, plain, devoid of added things. You are using a kernel customized for xen, which you MAY want to do, but if you are not using xen currently, with only the possibility that you may wish to add it, I would boot a different kernel until such time that you really require xen (though I like KVM or VirtualBox now).

i`m aware that vanilla kernel , is de-voided from added stuff , but i didn`t try a pure kernel before , as i told u still newbie:p

thats was a great add to mention , i`ll try out these tools , thanks a lot brother ..
one last thing please , is it much difference between Raid1 and Raid1E,
i know Raid1 require 2 disks and u lose 50% of space , but how Raid1E differs ?
thanks again ..