Hi Susan and Rene, I am using exactly the same scheme in my DAQ and it works quite well. When you use a root file/tree as an event buffer, you do not need to write a complicated memory based event or histogramming service. A user can write a root script to display and update histograms (near on-line) he/she needs with any additional features, etc. I am using a lock file as Rene has suggested and I would say the probability of collisions is not small, in fact you may not notice any consequence of some of them, but some of them kill the system quite easily. I am working with an intermediate energy experiment where number of collisions is not one per day, so that as far as you above 100 events per seconds you will die quite soon without resolving the collisions... The lock file scheme is far from being optimal and it slows down the performance. Rene, could you think a bit about TFile and TTree and suggest more elegant solution? It looks like you will make two groups happy at least :) Regards, Anton
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:35 MET