The touch count reset has significant ramifications. First, this means touch counts will not skyrocket into infinity. Figure 6-21 is an example of what you may find in your production system. It may be surprising, but due to the touch count reset during promotion, buffer headers typically have a touch count of 0, 1, or 2. The touch count reset also means the most popular buffer headers will not necessarily have the highest touch counts. If you notice a particular buffer has a low touch count, you may have caught a popular buffer just after it was promoted or crossed into the cold region. In fact, the highest touch count buffer headers will reside near the LRU end of the LRU chain!

With all this touch count incrementing, resetting to zero, and touch count inquiry, the definition of a popular buffer begins to get a little nebulous. Is a popular buffer one that remains in the cache for 8 hours with an average touch count of 2, or a buffer that remains in the cache for 5 minutes with an average touch count of 250? Well, Oracle has determined a popular buffer simply needs a touch count greater than or equal to 1 (by default) when asked.

Figure 6-21. Shown are touch counts followed by the number of buffers with that touch count (the number of occurrences). Notice that the touch counts are much lower than you may initially think. This is due to the touch counts being reset.