Notes on monitor program
~~~~~~~~~~~~~~~~~~~~~~~~

All code can be found in 
	/afs/cs.cmu.edu/project/cmcl-darwin/kernel/app.common/


monitor:
	To run: monitor interface_name linkspeed (in bits/ms)
		e.g. monitor de0 100000
	monitor has to be run on the router machines. 


monitor.remote:
	Can be run on a remote machine. See README.* for detail or 
	contact qihe@cs.cmu.edu

   	
Example of adding a class:
	Suppose you want to create a class at the output interface de0 
	on router1:

		router1 de0 -------> router2

	(1) run "monitor de0 100000" on router1, then press the "Add" button.

	You need to supply two things:
	(a) filter: src ip prefix, dst ip prefix, proto, src port, dst port
	(b) service curve: m1, m2, d

	A filter describes the set of packets that belongs to a flow.
	We use TCP/IP header fields. Src IP and Dst IP can be prefixes,
	expressed in an address and a mask pair. For example, 128.2.*.*
	is represented as address "128.2.0.0" and mask "255.255.0.0".
	protocol number, src and dst port numbers are _not_ prefixes.
	You can however specify zero for each of these fields if you 
	want them to be "don't cares".

	There are two types of service curves, concave and convex:


	bit	|m1    m2			|
		|  /-------			|      /
		| /|				|     / m2
		|/ |				|m1=0/
		|-------------			|-------------
		   d		time		    d

		concave				convex

	Restriction -- concave: m1 > m2, convex: m1 = 0

	m1&d specifies the delay experienced by a transfer unit of size (m1*d)
	m2   specifies the long term overall throughput guaranteed.

	concave : gives you lower delay, good for realtime apps (voice)
	convex  : gives you higher delay, but still good throughput,
	          good for ftp

	See the H-FSC paper for detail on service curves and the algorithm.

	Admission control:
	A request to add a node with a particular service curve may or may
	not be admitted, the parent node in the tree has to have enough 
	resources to support all the children nodes in the tree.
	A node is admitted when the sum of all the children service curves
	does not exceed the parent service curve at all time.

	(2) Press "Add". A new node appears in the main window.

	(3) You can now click on the newly created node and a window 
	will pop up with the information associated with this node.

	(4) You can modify the node's filter and service curve using the
	corresponding "modify" buttons.

	(5) You can delete the node using the "delete" button.

	(6) You can use the monitor to obtain visual performance feedback:
		(a) Start some traffic from router1 to router2 through de0.
		(b) Press "Monitor On" button in the main window.
		(c) Wait a few seconds for the data to be collected.
		(d) Press "Monitor Off" button in the main window.
		(e) Click on the node that you are interested in getting
		    performance feedback from, a window pops up.
		(f) Click on "Display Plot" button.
		(g) Two graphs will show up if there are data collected for
		    this node.
		(h) In delay plot, each bar is a packet, the height 
		    is the queueing delay experienced by that packet. 
		    All packets that were classified to this node
		    or any nodes below this node is displayed.
		(i) In the bandwidth plot, the bandwidth sharing of this node
		    and the immediate children of this node is color coded
                    in the plot. Brown is the node itself, other colors
                    represent the children nodes, the mapping of the
		    colors is the same as the link colors used in the 
		    main window. The sharing is expressed in percentage.

	(7) To quit, press "Quit".

	(8) Warning: DO NOT use "Backlog On".