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

 
Thread Tools Search this Thread
Operating Systems Linux Red Hat Is overlapping two RAID5 arrays on same drives a bad idea ??
# 1  
Old 06-04-2011
Error 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?

Code:
            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

# 2  
Old 06-04-2011
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.
# 3  
Old 06-04-2011
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 ..

Last edited by Saed; 06-04-2011 at 04:35 PM..
# 4  
Old 06-04-2011
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.
# 5  
Old 06-04-2011
Error

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

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 .
Quote:
(chunk size should depend on the size of the majority of your files and the type of file system.)

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
..
# 6  
Old 06-06-2011
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).
# 7  
Old 06-07-2011
Quote:
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.
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 newbieSmilie
Quote:
I tested with things like fio, and bonnie++ to get results with the devices
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 ..



Login or Register to Ask a Question

Previous Thread | Next Thread

9 More Discussions You Might Find Interesting

1. Shell Programming and Scripting

Identify the overlapping and non overlapping regions

file1 chr pos1 pos2 pos3 pos4 1)chr1 1000 2000 3000 4000 2)chr1 1380 1480 6800 7800 3)chr1 6700 7700 1200 2200 4)chr2 8500 9500 5670 6670 file2 chr pos1 pos2 pos3 pos4 1)chr2 8500 9500 5000 6000 2)chr1 6700 7700 1200 2200 3)chr1 1380 1480 6700 7700 4)chr1 1000 2000 4900 5900 I... (2 Replies)
Discussion started by: data_miner
2 Replies

2. SuSE

Raid5

Hi all, I am currently using opensuse 12.1, We have Raid 5 array of 8 disks. A friend of mine accidently removed a drive & place it back and also added a new disk to it(making it 9 disks). now the output of mdadm --detail is as shown below si64:/dev # mdadm --detail /dev/md3 /dev/md3:... (1 Reply)
Discussion started by: patilrajashekar
1 Replies

3. What is on Your Mind?

Idea: selling Puppy Linux flash drives

What do you think of the idea of selling flash drives already prepared to boot Puppy Linux on PCs? (I still need to see if I can boot on Macs using online instructions I've found.) I know it's not too hard to prep your own flash drive, even if you have to buy one first, but just think about why... (8 Replies)
Discussion started by: MrMormon
8 Replies

4. Programming

question about int arrays and file pointer arrays

if i declare both but don't input any variables what values will the int array and file pointer array have on default, and if i want to reset any of the elements of both arrays to default, should i just set it to 0 or NULL or what? (1 Reply)
Discussion started by: omega666
1 Replies

5. OS X (Apple)

Deleting a recursive symbolic link was a very bad idea

Well i was tidying up some files in a very important directory on our development server and somehow some plank had put a recursive sybmolic link in it. Which I the even bigger plank tried to delete from my FTP client. My FTP client then thought it would be OK to delete not only the sybmolic link... (0 Replies)
Discussion started by: timgolding
0 Replies

6. Shell Programming and Scripting

how floppy disks, CDs and flash drives (pen drives) are accessed in UNIX

hi how floppy disks, CDs and flash drives (pen drives) are accessed in UNIX? thanks (0 Replies)
Discussion started by: nokia1100
0 Replies

7. Shell Programming and Scripting

using case to do this might be a bad idea?

Reading this file. I want to read all 4 fields. If 2 is populated with a p, I want to set $TYPEP to "Printers", if not should be empty. If 3 is populated with an r, I want to set $TYPER to "REQ Printers" if not should be empty. If 4 is populated with letter o, I want to set $TYPEO to... (12 Replies)
Discussion started by: Skyybugg
12 Replies

8. UNIX for Dummies Questions & Answers

Why is it Bad Idea to insert "." (Dot) to PATH ?

I was told that it's a Bad Idea (especially for root ) to Add To the Variable $PATH in unix the ":." (dot), In order to execute programs in my current directory without typing ./program For example: PATH=$PATH:$HOME/bin:. Does someone know why is it a Bad Idea? (2 Replies)
Discussion started by: amitbern
2 Replies

9. Shell Programming and Scripting

Limitations of awk? Good idea? Bad idea?

Keeping in mind that I'm relatively comfortable with programming in general but very new to unix and korn/bourne shell scripts.. I'm using awk on a CSV file, and then performing calculations and operations on specific fields within specific records. The CSV file I'm working with has about 600... (2 Replies)
Discussion started by: yongho
2 Replies
Login or Register to Ask a Question