Striping mechanism

ITEM: RS4000005627



QUESTION :                                                                      
Hi,                                                                             
I need some clarifications concerning the striping, as managed by AIX.          
If I have to stripe 33 Kb of data, and if the stripe unit is 8Kb, on            
4 disks, the first 8 Kb are stored on the first disk, the 8 next on the         
second, and so on , but what happens for the last 1K ?                          
Are they stored immediately or do the system wait to have a complete            
block of 8 K to write ?                                                         
On which disk is it written ?                                                   
( I doubt that it is written on the first disk, but the iostat shows            
more activity on the fourth disk )                                              
Does the type of disk ( SCSI, SSA ) have an influence on the mechanism          
used by AIX to strip ? ( Is it the same algorythm which is used in both         
cases ?                                                                         
Please could you indicate a documentation where I could get more               
informations concerning the striping ?                                          
Thank you                                                                       
                                                                                
ANSWER :                                                                        
                                                                                
If the stripe width is 8Kb, then in the scenario you describe, the last         
1Kb of data will be written to the first hdisk:                                 
                                                                                
       hdisk1    hdisk2    hdisk3    hdisk4                                     
     ---------------------------------------                                    
    ! 1st 8Kb ! 2nd 8Kb ! 3rd 8Kb ! 4th 8Kb !                                   
     ---------------------------------------                                    
    ! 5th 8Kb ! ...                    !                                        
     ---------------------------------------                                    
                                                                               
The algorithm is independent of the type of disk subsystem used.                
                                                                                
In general, the system writes/reads in 4Kb pages, with a blocksize of           
512 bytes.  This means that the system will write out 4Kb at a time.            
The remaining 1Kb is buffered until the syncd daemon flushes the buffer         
(usually every 60 seconds or so) or until another 4Kb page is written.          
This is normal for AIX, and is not a feature or by-product of                   
mirroring.                                                                      
                                                                                
For additional resources, consult the System Management Guide, the              
Performance Tuning Guide, or the IBM Red Book on AIX Storage                    
Management.                                                                     
                                                                                
                                                                                
                                                                               


WWQA: ITEM: RS4000005627 ITEM: RS4000005627
Dated: 08/1996 Category: AIXSTORAGE
This HTML file was generated 99/06/24~12:43:04
Comments or suggestions? Contact us