Funny hdd results

[unijab]

Guess I should make my writing and reading test larger than the amount of ram available. It seems as though all my read speeds were affected by caching, because Im sure I shouldn't have read speeds that high. I also have mixed results with the write test

 

 

[Gilbert_Sarip]

what os are you using? 
what command is 'speed'?

[TheGeekster]

That's a bit, vague, what system are using? As Gilbert_Sarip asked, what OS? What hardware? etc etc

[unijab]

CentOS 6.6 

6x WD Red 1TB 

 

 

speed is a bash script i made. It writes and reads using dd command, then deletes the file.

[Speedbird]

what os are you using? 

what command is 'speed'?

 

 

That's a bit, vague, what system are using? As Gilbert_Sarip asked, what OS? What hardware? etc etc

 

 

CentOS 6.6 

6x WD Red 1TB 

 

 

speed is a bash script i made. It writes and reads using dd command, then deletes the file.

Quote posts, please.

 

And yeah, those results are a bit unusual, especially for a 5400 rpm hard drive. Could be some kind of RAM caching.

[CWP]

CentOS 6.6 

6x WD Red 1TB 

 

 

speed is a bash script i made. It writes and reads using dd command, then deletes the file.

 

You will need to use sync to ensure that data written is actually flushed to the disk and not stored in RAM. Add either one of the following to dd:

• conv=fdatasync: Ensures that a sync is done at the end of the entire write session (ie once)
• oflag=dsync: Ensures that a sync is done at the end of each block size (ie. count= times)

For sequential read tests, look into hdparm -t /dev/.... However, it is not installed by default (or at least from the minimal install). If you want to stick with dd, you may want to consider randomizing the start position of the read (skip= which also counts in block size chunks).

[unijab]

First 4 are writes;

Last 4 are reads

 

Varying blocksizes from 1M to 4M, increments of 1M.