From: anton.txt@g{oogle}mail.com   
      
   I wrote:   
      
   > We have a server with usually seven to ten active MSSQL   
   > instances, of which only few are intensively used at each   
   > moment.  Manually stopping unused instances and starting   
   > required ones several times a week is too burdensome for   
   > several reasons.  In order to improve RAM utilisation, I   
   > have written a simple script that tries to decrease memory   
   > use by instances not in active operation.   
   > [...]   
      
   I continue to experiment with my memory balancer, and its   
   worst defect is that sometimes by decreasing   
   'max server memory' it causes the isntance to freeze and   
   become unresponsive until restarted.  There are three ranges   
   of these setting:   
      
           freeze    high CPU      Works OK, zero CPU   
           until     usage while   usage while idle   
           restart   idle   
          |--------|-------------|--------------------->   
          0        V1            V2   max server memory   
      
   where V1 and V2 are idiosyncratic to each instance.  I have   
   tried to estimate V1 and V2 using:   
      
   I have tried monitoring the following parameters:   
      
      1. sys.dm_os_process_memory:   
           - process_physical_memory_low   
           - process_virtual_memory_low   
      
      2. sys.dm_os_performance_counters:   
           - Memory Grants Pending   
      
      3. sys.dm_os_memory_pools:   
           - free_entires_count   
      
      4. a percentage of the total .mdf file size in the instance   
      
   but in vain.   
      
   Have you an idea how to estimate minimum working value for   
   server memory, that is V2 in the diagram above?   
      
   --   
   ()  ascii ribbon campaign - against html e-mail   
   /\  http://preview.tinyurl.com/qcy6mjc [archived]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)