Update README.md

@@ -27,7 +27,25 @@ server always needs all available cpu resources given to the machine, minimizing
 the risk of letting the hypervisor think that cpu resources can be shared with
 the risk of letting the hypervisor think that cpu resources can be shared with
 some other virtual machines. This helps in for example performance testing,
 some other virtual machines. This helps in for example performance testing,
 where it is important to have the same amount of resources available each time
 where it is important to have the same amount of resources available each time
-a performance test is run.
+a performance test is run, in order to compare with previous results in the same
+machine.
+
+It is of corse better to change the virtual machines resources settings to get 
+dedicated CPU resources and high priority, but this is not always easy to
+achieve, since some virtualization setups want to have the same settings on all
+machines and let the hypervisor decide where to put what virtual machines. The
+nature of performance testing, where one hour of the day can require all
+available CPU resources and the rest of the day does not use any CPU at all, is
+extremly difficult for the hypervisor to predict, and provide addequate CPU 
+resources when needed during a test.
+
+For this reason, it is good practice to keep the piloop tool running a day or
+two before the performance testing starts, in order for the hypervisor to make 
+room and provide CPU resources for all virtual CPU's. In order to not affect 
+other virtual machines by suddenly requireing all CPU's, its also good practice
+to more slowly increase the number of CPU's being fully loaded with piloop. You
+can start several piloop processes with a couple of hours in between, where each
+piloop process loads 25-50% of the available logical CPU's.
 
 
 ## Compile
 ## Compile