From: Marcos D. de A. <ma...@cs...> - 2007-09-24 10:27:08
|
Hi Anthony, thanks for your email. I just want to understand to check if I am not doing anything wrong by ignoring the conversions. > For example, the resource is in UTC+10 and the user is in UTC+5 > and init time is 05:00:00 (UTC+0) for simplicity. > At the start of simulation, resource time is 15:00:00 and > user time is 00:00:00 > If the user submits at GridSim.clock() = 1 hour and > NO time zone conversion, then the resource thinks > the current time is 01:00:00, i.e. user init time + GridSim.clock(), > which is already expired. This is because the ARObject only stores > the user current time and the user has no idea about > the resource time zone is. I do agree that in a real distributed application the time zones have to be handled. However, in the simulation, the user will make a reservation request relative to the simulation clock (in seconds). The provider will abide to the same clock to check whether it accepts or rejects the request (in seconds). If a user requests 5 PEs at simulation time 60 sec. for 1 hour, then the provider will check if there are 5 PEs available at simulation time 60 sec. for one hour. The ARObject may handle the times as doubles reflecting the simulation clock. I do not see the need for time zone here. However, let's suppose that the provider has a policy that states that resources may not be reserved on Mondays. The provider has information in which time zone it is. In this case I see the need for converting the 60 sec. simulation time to the provider's time zone to check whether it is a Monday or not. Apart from cases like this one, I do not see a clear reason for using relative times. I may be wrong. Regards, Marcos -- Marcos Dias de Assuncao Grid Computing and Distributed Systems (GRIDS) Laboratory Department of Computer Science and Software Engineering The University of Melbourne, Australia Email: ma...@cs... |