Most Important First (MIF) Scheduler
====================================

Table of contents
-----------------
1. Introduction
2. Conf file parameters
3. Running the example.

1. Introduction
---------------

This scheduler manages a ready queue of Distributable Threads (DTs)
ordered by their importance and suspended on a condition variable. AT
any given time the scheduler dequeues the DT at the head of the queue,
that corresponds to the most important thread, and signals the thread
to activate it. The service context is used to send the importance and
GUID of the DT across hosts it spans so the DT can be scheduled
accordingly on the remote host.

In this experiment we show how dynamic scheduling is done using the
Dynamic Scheduling framework with the MIF Scheduler as the pluggable
scheduler. At any given instance the DT of highest priority is running
on a given host.

The experiment consists of the following participants:
Job: A CORBA servant object that performs CPU intensive work. The
amount of work depends on a load factor that is conveyed to the object
per invocation as an argument.

DT_Task: The distributable thread is a scheduling segment spanning one
or more hosts.

test: The test consists of a collection of Jobs and DTs hosted in
a single process. The test reads a configuration file that can be used
to initialize DTs and Jobs.

Starter: The starter initiates the start of DTs on each host that the
test is running on. This is to ensure that the experiment starts at
the same time on all hosts.

2. Conf file parameters
--------------------

POA OPTIONS
==========
To specify POA options the format is:

-POACount <count> -POA <name> -PriorityModel <CLIENT|SERVER> <priority> -Lanes <count> (-Lane <priority> ,<static_threads> <dynamic_threads>)* -Bands <count> (-Band <low> <high>)*

e.g.
-POACount 2 -POA poa1 -PriorityModel CLIENT 10 -Bands 2 -Band 1 20 -Band 30 85 -Lanes 2 -Lane 10 1 0 -Lane 80 1 0

specifes a POA with:

POA Count - Specifies the number of POAs that need to be activated. The following characteristics of each POA needs to be specified.

Name - poa1

Priority model - client propogated, def. priority = 10

Bands - 2 Bands with Band values as follows -
	Band 1 :  low priority = 1, high priority = 20
	Band 2 :  low priority = 30, high priority = 85

Lanes - 2 Lanes with Lane values as follows -
	Lane 1 : priority = 10, 1 static thread, 0 dynamic threads
	Lane 2 : priority = 80, 1 static thread, 0 dynamic threads


Distributable Thread Task Options
=================================

The format to specify a DT Task is:

-DT_Count <count> -DT_Task Importance <imp> -Start_Time <time> -Iter <local_work> -Load <remote_work> -JobName <name>

where,
-DT_Count = Total number of DT_Tasks
-DT_Task = Specifies a Distributable Thread
-Importance = The priority of the DT
-Start_Time = Time at which the DT enters the system
-Iter = The number of secs of work to be done on the local host. For a
distributed DT it defines the number of secs of local work to be done
before and after a remote method call is made.
-Load = The number of secs of work to be done on the remote host when
a two-way method call is made by a distributed DT
-JobName <name> = Name of the Job object that this DT will make a
remote method call on to do some 'Load' number of secs of work on the
remote host

e.g.
-DT_Count 1  -DT_Task -Importance 5 -Start_Time 0 -Iter 3 -Load 5 -JobName job_1

specifes a DT Task in which,
Importance = 5
Start Time = 0
Iter = 3
Load = 5
JobName = job_1

Job Options
===========
The format for specify a Job is:
-Job_Count <count> -Job <name> <poa_name>

where, poa_name is the POA that this object is activated in.

Job Count - Specifies the number of jobs that are activated.

e.g.
-Job job_10 poa1

specifies a Job with,

Name - job_10
POA Name - poa1

Misc
=====

-GuidSeed <guid>

This specifies the guid number with which the guid counter will get
initialized. This is to ensure that unique guids are used. This will
be removed when the ACE UUID generator is integrated with TAO.

-OutFile <filename>

This specifies the data file in which the schedule of the DT"s running
on the host will be output.

-LogFile <filenmae>

This speficies the log file

3. Running the example
----------------------

a) The activated Jobs and Synch objects are registered with a Naming
service, so we need an NS running

e.g. ./Naming_Service -o naming_ior

b) Start one or more instances of ./test depending on the test
configuration that you have designed.

e.g. ./test -ORBInitRef NameService=file://naming_ior -ORBSvcConf svc.conf.whatever -ORBDebugLevel 1

c) Execute the Starter that initiates the creation of DTs on all the
hosts that the experiment is running

eg. ./Starter -ORBInitRef NameService=file://naming.ior

c) Once all the instances exit, the test will generate schedule files
as specified by the user with the -OutFile option file