Meeting Minutes:  Team Meeting w/ Priya
Time: 04/14/04 @ 6:00PM-6:30PM
Where: MSE Cave
Attendees: Ackley, Boyer, Fry, Wilson
Purpose: Weekly Status Meeting
------------------------------

--- Presented data collected from last week ---

*) Largely, data focused on how we reduced the bounded turnaround time 
   in the presence of faults, and the system behavior for fixed vs. 
   variable sized messages.

Note: A bug was uncovered this past weekend where some of the graph data
      was not being properly seperated.  Something to keep in mind when
      analyzing graph data.

Priya's Comments:
*) Look on google for the linux scheduling algorithm.
   - spend no more than 10 minutes to find support for theory that 
     linux scheduler is impacted more by positive load delta than by
     constant load values.

*) Could use a correlation algorithm between msgsize and response time
   to try to predict what respose time will be based on msgsize.
   - this is optional and not required for this course.

*) Multi-client data supports theory that positive load delta's 
   introduce significant reponse latency.  More-so than a steady
   increased replica load.


*) For next week, implement the load balancing strategy and show how
   this improves the reponse time for muptile clients.
   - Focus is on total client response time.
   - For example:

	     ^                        ^
    response |	   /		      |               /    with load
    time     |    /   No load         |             --     balancing
	     |   /    balancing       |------------/
	     |--/		      |
	     |			      |
	     +------------->	      +---------------------->
	        # clients                # clients
   
   - expected results:
      A) there is a penalty to implement load balancing, which is
         shown by an overall latency increase (upward shift) even under 
         light load conditions.
      B) benefit is that load balancing should perform better under
         'heavy' loads than without it, as shown by being able to
         handle more clients more efficiently than otherwise would
         be possible (shown by shift to the right)

*) No class on this Thursday

*) We have opted for earliest presentation time.
   - We will have dry run on Monday.
   TODO: We need to notify Priya of available meeting times on Monday
         to give a dry-run of the presentation.
   - If there is a problem meeting this schedule, we should notify
     Priya as early as possible to determine action to take.


------------------------------------
Team Meeting:
Time: 04/14/04 @ 5:00PM-6:00PM
Attendees: Ackley, Boyer, Fry, Wilson
Where: NSH Atrium
Purpose: Prepare for Meeting w/ Priya

Agenda:
*) Review data from past Saturday
*) Review items to present to Priya
*) Update plans for upcoming week.


Group Plan: To complete load balancing strategy by Saturday, and to
   begin collecting load-balancing data.


New Assignments:
----------------
GA/PW - Implement Load balancing for Saturday

AB - Collect data using new load balancing strategy
     - variables:
       - fixed vs. variable msgs
       - number of clients
       - number of replicas

?? - Implement multi-threaded mechanisms in EatSpam client to:
     A) improve performance, simulating a multi-threaded MTA
        which can simultaneously process many messages
     B) help ensure bounded failover for multiple simultaneous
        message processing requests (real-time)

?? - Prepare slides for presentation on Monday.