aglew@ccvaxa.UUCP (06/06/86)
Somebody just came into my office wondering about tuning the buffer cache. Seems he had been working since early morning, and noticed the performance fall off as more people logged in to the system; he also noticed that the system was certainly not compute bound, and didn't seem to be doing too much I/O. He wonders if the buffer cache might be thrashing. Does anyone have any information, intuition, experience, or references to studies with numbers, about tuning buffer caching? What are typical values for buffer lifetimes, buffer reuse, hit ratio, etc., at different load levels? Has anyone tried working set style strategies for buffer caching? Are there any particular reasons why it wouldn't work well? Please send mail; I'll summarize to the net. Andy "Krazy" Glew. Gould CSD-Urbana. USEnet: ihnp4!uiucdcs!ccvaxa!aglew 1101 E. University, Urbana, IL 61801 ARPAnet: aglew@gswd-vms