A while back, when we were creating our Linux version of CounterSpy (for use on gateway appliances), we needed to optimize CounterSpy’s large database for highly efficient, high-volume use.
One of our senior developers, Chandra Prakash, developed a modification of the classic B-tree data structure and the results have been quite impressive in terms of performance. He’s presenting the methodology at the AVAR conference this week.
Being the kind and beneficent soul that I am, I’ve posted his paper here.
After all, I know my competition needs all the help it can get.*