Meeting Minutes


Purpose: Project Proposal – Meet Cody in person

Date: 9/13/07
Time: 2:00-3:00 PM
Location: Dr. Junk’s Office JEB 324

Attendees: Dr. Junk, Cody, Frank, Michael, Ryan, and Andrew

Recorder: Andrew



Summary of Items Discussed:

2:00 PM – 

Ryan presented project proposal-

 -Slides to presentation are on website: www2.cs.uidaho.edu/~cs481-50/
 -Talked about SEL logo
 -Talked about customer motivation:
    -We’re looking for Cell processor utilization/capabilities in this project
 -Talked about product solution highlights:
    -We’re looking at investigation of Cell, not research: THERE IS A DIFFERENCE!
    -Talked about software releases: when/what/how?
 -Talked about use cases:
    -Talked about benchmarking
    -Cody suggested funnel in and out applications with wrapper script
    -Talked about output files / we should make files universal to work with multiple  
      programs
    -Cody said we should focus on intricate parts of algorithms 
 -Talked about architectural approach:
    -Input -> command line interface
    -Tie in benchmarking
    -Cody said we should build interface and establish set of “golden files” – good for  
      development
    -Cody talked about building hardware encryption board – Problems were found 
      because good benchmarks and regression testing were not implemented correctly
    -Cody said we need to look for false positives and negatives
    -Ryan looked at Cell architecture and described how we want to implement AES
    -Cody said AES hard to accelerate with networking on Beowulf clusters but with Cell 
      processor there it should significant improvement.
    -Cody said we might want to look into streaming side of AES.
 -Talked about testing strategies:
    -Dr. Junk said we should test front end such as command line parameters.  Make sure 
      that inputs produce expected results.  Ex:  Check for bad inputs.
    -Need to test for correctness according to application specifications
    -Discussed how benchmarks are really a subset of tests that show whether program 
      meets specifications
 -We looked at project schedule:
    -Dr. Junk said there needs to be a code freeze 5 days before each release.
    -Need documentation/presentation prepared for each release.
    -Release 3 might have another strategy such as doing a streaming version of AES
    -We would probably learn more by implementing another strategy rather than just 
      optimizing what we have done.
    -Talked about possibility of complier bugs
 -Talked about risk analysis:
    -Compiler bugs
    -We could get in too deep with optimization and screw things up
    -O/S stability might be a problem
    -What if someone gets sick? – Other team members take over
    -PS3 dies
    -We can use GDB kernel debugger if some PS3 hardware issue comes up
 -Asked Cody if he had any questions:
    -How comfortable is group?  We are comfortable with project.
    -Scope?  It looks fine.
    -What kind of throughput results have we gotten? – We need to try ISO image.
    -Cody said typically you get better results from specialized hardware than mainstream 
      with specialized applications.
 -Frank asked Cody how he would do streaming mode.
 -We need a parallel decryption scheme as well
-Cody suggested once again a wrapper script around applications.
-Talked about definition of latency- Either time from first block in to first block out
                                                          Or       time from first block in to last block out
-Can use built in time function if doing second option
-Cody said latency and throughput are good measures to start with.
-Dr. Junk thought presentation went well.  He said it appears we are making good progress.
-1st release of code should show significantly if we are on the right track
-2nd release of code will probably be the most significant release of the project.
     
3:00 PM-

-END-


Action Items:

Set up our own CVS repository
Work on release 1
Work on documentation for release 1
Work on legal/ethical/social review
Need to get SEL logo
We need to encrypt ISO images and other large files to build up test bank.
Look into how we might implement a streaming version of AES
Figure out how we are going to benchmark original AES code.
Figure out encryption/decryption of parallel AES.