From: Giuseppe Mornacchi [Giuseppe.Mornacchi@cern.ch]
Sent: Friday, March 28, 2003 1:40 PM
To: Wickens, FJ (Fred) ; serguei.kolos@cern.ch;
beniamino.di.girolamo@cern.ch; Benedetto Gorini; reiner.hauser@cern.ch;
bob.dobinson@cern.ch; saul.gonzalez@cern.ch; fred.wickens@cern.ch;
david.francis@cern.ch
Cc: giuseppe.mornacchi@cern.ch
Subject: varying assumtpions..


 Please find enclosed contributions from Bob and myself; no other has
arrived yet...
 giuseppe

Bob
============================================================================
============
Amount of data that needs to be collected by level 2 is larger than
foreseen

LVL2 rejection is lower than expected leading to larger event building
rate requirements

Hot spots in the ROBs leads to congestion in the ROS and access network

Event farm size unknown because of higher event building rate is in any
case larger than expected.

Monitoring requirements exceed estimates

============================================================================
============
giuseppe

 Relaxing assumtpions & possible effects on baseline


 1) Level 1 rate: now target is 75 KHz, level-1 may run up to 100 KHz
    		  Implication... possibly everywhere

 2) ROI Size: assumption is 2%. WHat if larger than 2%; how much larger
could it be?
		  Implication.. ROL mux factor into ROB; ROB ouput; ROB
mux/concentration;..

 3) ROI request rate/ROB: the current assumption is 2% on average? some ROBs
may be hit more (how much)
			  frequently than others. Depends on detector? on trigger type?
		   Implications: same as 3) probably

 4) Level-2 rejection: today assumption is a rejection power ~ factor of 30.
What if it is not reached?
		   Implications: higher rate into EB, then: ROL/ROB mux/concentration
factors; number of SFIs;
	           size of EB network

 5) Level-2: what if (may beunrealistic, but should we cosider the
scenario?) level-2 shows to be not
	     cost effective?

 6) Hz/L2PU (events a L2 box may handle/sec): today assumption is ~500.
What-if it is less? (by how much?)
		   Implications: size of Level-2 farm, hence level-2 network.

 7) SFI rate IN: how many events an SFI can build/sec. Today assumption is
75 MB/sec. Is there a need for a what-if?

 8) SFI rate OUT: how many events/sec an SFI can serve to the EF?

 9) Hz/EFPU: I do not know what the assumption is (1HZ???), we should study
different scenarios
		Implications: could it have implications, a part from the obvious one on
the size of the
		EF and its network, on the number of SFI?

10) There is the question of the event size, but here we have probably
already been quite conservative.