Hash: SHA1

In working on VCL-16, I realized that the time table that gets presented to a 
user when a requested environment is not available only works correctly for 
single images.  If it is presented for a cluster reservation, only the parent 
image is used to calculate availability.  Because the child images could have 
different hardware dependencies than the parent, either a separate time table 
would need to be generated for each image, or a collapsed version of a time 
table would need to be generated that displays times in a single column that 
the whole cluster would be available.  The former doesn't scale at all for 
larger clusters (imagine working with 10 time tables for a cluster with 10 
images).  The latter would be difficult to generate.

So, for all reservations (single or cluster), I'm thinking of replacing the 
time table idea with an interactive "auto find" method.  This method would 
iteratively try to find a time the request would fit in by keeping the 
requested length the same but increasing the requested start time by X units 
of time (X would probably increase relative to the requested length of the 
reservation).  I'm thinking to do a small number of iterations (Y).  Then, if 
one or more available times were found, prompt the user to see if any of them 
would work.  If no times were found or if the user didn't like any of the 
listed times, the user could click a button to have VCL iterate another Y 
times and repeat the cycle.

The other benefit of this is that it would introduce the "packing problem" in 
to VCL providing more thesis topics for grad students. ;)

Any thoughts?  Can anyone think of a way to keep the time table without too 
much difficulty?

- -------------------------------
Josh Thompson
Systems Programmer
Virtual Computing Lab (VCL)
North Carolina State University


my GPG/PGP key can be found at pgp.mit.edu
Version: GnuPG v1.4.6 (GNU/Linux)


Reply via email to