With multiple real processors, it is certainly possible. We have the same potential problem and plan a circumvention of adding (or multiplying) the server number to the random result:
If servers are named xxxnnn, random() + right(userid(), 3) shoud be unique. Of course it depends how large a range you are requesting. Random(0,1) has a 50% chance of a collision. A veriation of that may be what you are experiencing. -----Original Message----- From: CMSTSO Pipelines Discussion List [mailto:[email protected]] On Behalf Of Bob Cronin Sent: Thursday, February 17, 2011 11:42 AM To: [email protected] Subject: [CMS-PIPELINES] How random is the random stage? If "n" virtual machines all running the same code all invoke a PIPE command that uses the random stage without a seed specified and they all do it at very nearly the same instant in time, will they all generate the same sequence of random numbers? I have evidence that suggests this is true (and if so, I need to come up with a scheme to add a seed that will be different for each virtual machine). -- bc
