|
Home > Archive > Server + > August 2002 > Which one has better performance raid0+1 or raid 5?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Which one has better performance raid0+1 or raid 5?
|
|
| twister166 2002-08-20, 12:59 pm |
| I am having a difference of opinion here:
I think Raid 5 is fater than raid 0+1, my buddy disagree. I need judges!
My resaon for Raid 5 is faster, because that data will only require 1/n times to write the parity and only writing parity.
While 0+1 is not writing paraity but marror thus, the data amounts are doubled when it writes to the drive.
What I am tring to say if we would to write 30 MB of data to the hard drive:
Raid 5 will only write 30 MB of data plus parity information (small)
While, Raid 0+1 is stripping w/o parity but mirrow the data, it will 60 MB of data to write to drive.
Hence, less data to write, better performance!
I mean I don't have a text book, all I am doing is following the technical common sense...
 | |
| psnell 2002-08-20, 1:17 pm |
| Although it is hard to make a blanket statement about this. A lot of this is dependant on how well tuned the arrays are. Generally, however raid 0+1 will be faster than raid 5. This is due to raid 5 having to perform the checksum calculations combined with writing the parit information. I think an incorrect assumption is that with raid 0+1, you need to write twice the data when in fact the OS will just shoot the data to the controller then forget about it. The controller does all the work of mirroring, but the OS only sends it once.
This is my no means meant to be a definitive answer, I'm stretching the limits of my understanding on this one. | |
| twister166 2002-08-20, 1:31 pm |
| Under RAID 5, the OS only shoot the data once as well. Raid controller decide to do parity or mirrow based on your configuration of the RAID controller. It does not take much to generate parity, it takes alot to generate actual data from parity. It takes more to write to the drive than it is to calculate parity info. The average seek time is measured in Milli-seconds, the SCSI processor is measure in Nano-seconds.
Thanks for the input though... | |
| twister166 2002-08-20, 2:22 pm |
| Well, I went to Adaptec, based on their website, that O+1 or 10 is the highest performance. I can see it offers much more availablility to servers. It can sustain multiple hard drives failure. It may even offer a better reading throughtput, as it can read from the mirror set if the other is busy or failed. If the performance is based on reading, I will say the Raid 10 is better. But I am not convinced that when mirror writing is fater than parity calc + write...
Anyone has any suggestion as when I can read up? | |
| psnell 2002-08-20, 2:44 pm |
| try this
click | |
| Supertech 2002-08-20, 4:10 pm |
| When data is written to a RAID-5 array, the parity info must be updated. This is accomplished by finding out which data bits were changed by the write operation and then changing the corresponding parity bits in the following process:
1. Read the old data.
2. Write the new data.
3. XOR the old data with the new data.
The result is a bit mask which has a one in the position of every bit which has changed.
4. Read the old parity from the array.
5. XOR the bit mask with the old parity info. This results in the corresponding bits being changed in the parity info.
6. Write the updated parity back to the array.
Therefore, for every application write request, a RAID-5 array must perform two reads, two writes and two XOR operations to complete the original write operation. The cost of calculating and storing parity, rather than redundant data, is the extra time taken during write operations to regenerate the parity info. This additional time results in a degradation of write performance for RAID-5 arrays over RAID-1 arrays. A RAID-5 write operation is 33% to 60% slower than a RAID-1 write operation | |
| twister166 2002-08-20, 4:27 pm |
| Thanks SuperTech, that was the explaination I was waiting for. That make alot of sense for re-read before write.
Just now I have to buy my buddy a bear...
Damn, I don't like lossing!  | |
| Supertech 2002-08-22, 8:27 am |
| Why would you want to buy your friend a bear?
Is lossing kind of like flossing? | |
|
| mirror --> duplex --> RAID5 --> RAID0 from the slowest to the fastest disk configuration, all things being equal... you would have known that if you had read 70-215 for Freaks!  | |
| twister166 2002-08-22, 9:22 am |
| I mean beer not bear... sorry... | |
|
| LOL |
|
|
|
|