john@wvlpdp.uucp (10/01/86)
I have some questions about how to select the "proper" size of the disk buffer cache in the kernel. Are the following questions relevant: Is it possible for the buffer cache to be so large that the kernel uses more time looking for a block in the buffer cache than would be required to fetch it from the disk? If the buffer cache is "too big" would "too many" physical disk writes to be pending? If this is true would performance degrade, when the update daemon sync'ed(). If the above is true, would executing sync() more often improve performance? For our applications the size of the buffer cache has an impact on performance up to a point under a no load condition. Does a large buffer cache improve performance under a heavy load condition? We are using a PDP 11/73 running Ultrix-11 Version 3.0 1 Meg of memory and a rqdx1 controller and RD-52 disk drive (57.5 millisecond average access). Any comments on this subject are E-mail welcome. If I am way off base please politely point this out. Thanks in advance, John A. Ebersold World Video Library ihnp4!convex!ctvax!trsvax!hal6000!wvlpdp!john 2747 Airport Freeway Fort Worth, Texas 76111 (817)-831-3811