1a2
> 
10a12,24
> %Dennis,
> % I made a comment in the OLTP trace description that
> % our results are different because of different processing.
> % WRT the analysis, if the miss rate is small, (less that 10%), you
> % should use a smaller A2out queue, as the formula implies.
> % the reason -- if the miss rate is small there is little turnover
> % in the A2out queue, so you want to make it smaller to avoid
> % catching cold items.  This is implied by the formula.
> %
> % I've again posted the PS file to the ftp site if you want to see
> % the figures.
> %    Ted
> 
580,581d593
< {\em Ted: how did our results on the swiss page traces jive
< with those of Weikum?}
854a867,869
> We note that our results on this trace are different than those reported in
> \cite{OOW93} because we used a different method to process the trace file.
> However, our conclusions about relative performance are the same.
1151,1153c1166,1167
< {\em Ted: What if the miss rate is much smaller.
< Many administrators run their systems to have a miss rate less than
< 10\%, perhaps even less than 1\%}.
---
> For systems with a lower miss rate, the formula implies that a smaller
> A1out queue is better.
1155a1170
> 
1217a1233,1234
> 
> 
