Online organisation & costing. Notes for the meeting on 2 Apr 2003

											GMO
										     02-04-03


 - Saul, Benedetto, Haimo, Giuseppe


 -Open issue: disk vs diskless machines... ROS PCs, HLT PU etc.


  - ROD crates

	1 controller per ROD crate (the physical controller is as yet unspecified)
	There are O(100) ROD crates


 - ROS

	In the baseline there is a PC per ROS which can act as a controller
	There are O(150) ROSes

 - EB

	There are O(100) sFIs, each with a connection to the control network
	1 DFM

 - Level-2

	Use Fred's scheme:

	1 supervisory rack for RoIB and L2SVs
	~ 20 racks for the lvl2 processors (including switches etc.)
	1 controller per rack -> O(20) controllers

 - EF

	Use Fred's scheme:

	O(60) racks -> O(60) controllers (1 controller per rack)

 - Online systems

	This should include:

		"operator" consolles
		PCs to support the running of the online system
		monitoring farm

	in numbers enough so as to be partitionable


        The size of the "online system farm" should support the control, loading configuring and operational
	monitoring of partitions. We should guess what is required in terms of bandwidth for operations and
	operational monitoring. Information for event monitoring (mainly from RODs) is being collected by the
	monitoring wg; some reasonable guess should be made about the potential load of monitoring from ROSes
	and SFIs

 - Online (or control) network.

	The network could have ~ 500/600 nodes, but the number of potential sources of data (and data
	requests) is ~ 2000

	Should it be FE? GE? what topology?