Actually for the table format I suggested doing

look for the 'ended' value for try three of jobpid 200

would probably be faster than doing

WHERE try = 4

since the former is a primary key lookup (jobid & try) of the new table - where the latter might well have to do a full table scan. If you are doing queries like this a separate table for each try might be faster (although many databases would allow you to build a separate index on the try #).

:-)

That said I totally agree with Gyan's comments on the fact that DB design is not trivial. You know more about your data, and how it will be queried, so having a single table may not be the best solution.