Systems Research Laboratory

12:02 AM 4/8/2010

   All of this is going to yield a lot of data, even more, it is going to require a lot of execution time 
   to run cross products of all configurations.  

        A strategy for managing run times while carefully exercising workload granularity 
        variances will be necessary.  

   Warm up times may not be necessary as this is stateless system.
       But, once caching comes into play does this still apply?

6:26 PM 4/8/2010

   Experimental results for the SSM may not be completely fair as the experimental model I have designed 
   has introduced a flaw forcing 1/2 or more of a full stroke on the disk between the two workloads by virtue 
   of the file sizes on LUN layout I have chosen.

        I need to revise the model after presenting this.  Larger files will reduce the stroke length between 
        the two LUNs.

        This needs to be considered as a production class environment may over allocate capacity and hence 
        introduce track distances in a consolidated model/SSM.

        Outer Track of Disks |----------|
                             |          |
                             |          |<-- file on LUN 2 ends here
                             |----------|<-- file on LUN 2 is here and edges upward
                             |          |<-- extra seek distance between two random workloads (i.e. unused space on LUN 1)
                             |          |<-- file on LUN 1 ends here
        Inner Track of Disks |----------|<-- file on LUN 1 is here and edges upward

        This will make interesting presentation material.

12:17 PM 4/12/2010

   Does the information in the entry from 6:26 PM 4/8/2010 imply that file system fullness and file location 
   activity on the file system will affect the SSM's performance?