[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

No Subject




 > With most efforts to improve how things are done, it usually
 > pays to understand how it's done right now, to build on that.
 > How exactly does the "shuffling" process happen?  Does a guy
 > sort through them, and use his past experience to balance the
 > competing demands?

 > Maybe it could be improved, but if it's worked okay so far,
 > it must have some good points.


 The "agent" is a group of stored procedures in MSSQL 6.5.
 It has no memory of anything it has done.  The "Shuffling" is accomplished
 by assigning a random number to each record in the table, then
 running a query with a sort on that number.

 If the "agent" were handed a stack of 100 orders for the same commercial
break during the super bowl, after processing all 100 orders the only
commercials that would be placed would probably be the two or three with the
highest [$] and [Priority] properties. I suppose this is the best you can
ask for in that situation, but it may have taken the "Agent" an entire day
to do it.   I am hopeful that there is a cleaner (more scalable) solution to
this problem than merely patching up the major gaps in this "brute force"
method.

 The system has no problem with the amount of booking required for Radio.
Radio has very little traffic. The problem is with the Cable TV customers.
These customers are getting bigger and bigger and the amount traffic is
becoming such that the "agent" is treading furiously to stay just one day
ahead. These cable companies have thrown lots of money at the problem in the
form of hardware.

 With the new direction our company is taking, it's becoming clear that the
software will require scalability that is just not possible with the current
design.  I am working very hard to gather potential solutions to this
problem.

 I believe the solution may lie in the structures or processes found in
nature.

 The concept of the commercials competing to occupy the commercial break
spaces is an intriguing one.  : )
This "sounds" like it would be a GA type problem, but perhaps they're only
related in the semantic sense.


 Thanks,
Tommy Amabisca