Problem Statement:How to Optimize the IO's ...where through put is
                  degraded or not optimal where it has been clearly
                  been observed that the performance hindarance is
                  due to disk I/O's.

                  Ex: MailServers, database servers


Case-Scenario / realtime scenario
----------------------------------

Mailserver with a large number of mail box's operating on single disk.

Database servers with large number of I/O's happening on the same disk.


Due to crunch on the number of I/O's all stack of applications above
the disk layer are breaking or crashing.

How to fix or solve the application crash, Assuming that minimum amount
of modifications or no alternation's are required to tide over problem
or scenario.

options

a) Make use of the volume manager or disk manager ....spread the 

   IO's on to multiple disks ...use the technique of splicing or
   interleaves is used.

b) Given the filesystem is a hirearcial structure where disks or
   partitions can be mounted on to sub-directories.

   or mount the disk/disk-partition onto a sub-directory ...and
 
      virtually link the mounted partitions into the application
      directory structure.

   
   case-scenario: Had to manage a mail-server with all the data

                  of the users viz. the mailstore located under
                  the directory structure /mailstore.

                  and all the mailbox's of the users starting with
                  alphabet 'X' are organized under a sub folder
                  'X', all the mailboxe's of users starting with
                  'Y' are organized under the sub folder 'Y'

                 thus /mailstore/a
                      /mailstore/x
                      /mailstore/y ....lly respectively.

    Given the context where there are a high concentration of mailbox's

    under a specific alphabet ...the user's can be relocated to a
 
    seperate partion or mounted directory under the above structure

    viz. /mailstore/(~{?}=0) {...oops thus for anology sake kind of c++ virtual constructor}

Note: The above should come in handling or implement from large scale 
      to intermediate applications ...directory servers/ online
      resume database etc.

      One notable scenario or place would definitely organizing the
      directory or tabelspace structure under the MySQL DB.

      One drawback would be enable the applications to be able to
      handle the 'lost+found' directory that pop's up when mounting
      partitions.

      The following makes a innovative addition of tools to a system admin's

      swiss army tool kit ....additional options to the IOSTAT viz. observe

      the stats on virtual lun's and options to observe stats on the actual
 
      physical lun or disks when a raid layer is present at the OS layer or

      as part of the kernel in a operating system.


