From: "Robert Fox" <[EMAIL PROTECTED]> > On Fri, 2003-09-19 at 23:20, [EMAIL PROTECTED] wrote: > > On Fri, 19 Sep 2003, Robert Fox wrote: > > > > > I have three different machines running latest Cooker. I noticed that > > > the memory usage under the normal kernel (2.4.22-10mdk) seems normal, > > > but the reported usage using the big memory kernel > > > (2.4.22-10mdk-i686-up-4GB) is very large: > > > > > so? > > > > if you have lots of mem, it makes sense to cache all sorts of interesting > > data in your memory. Not to much so that if you start a program that you > > will have to free mem. But having your memory just sit there without data > > would be a waste. > > IMO, it is completely normal, and useful, behaviour. > > > > I want to agree with you - but 768 Meg of memory is still considered > "lots" - and if you look at the numbers carefully, you'll see that the > machine with 768 is showing proportionally more free memory as the > machine with 1280 Meg. >
If you look at the buffering, you see that the added 512MB between the two systems is almost a full match of the added buffering -/+ buffers/cache: 96 659 // 768MB system -/+ buffers/cache: 635 626 // 1280MB system ------------------------------------- -/+ buffers/cache: +539 -33 // difference wich adds up to 506MB added memory used for buffering... wich is exactly how the kernel is supposed to work... and giving the slight memory usage overhead of using highmem, since it need "lowmem" buffers to keep moving the data to/from highmem areas it's all ok... > Also, the machine with 1280 Meg showed much better memory usage when I > was running the earlier "Enterprise" kernels - before the i686-up-4GB > became available - so I suspect something is wrong here! If I recall, > it was showing under Gkrellm 1280M - 968M Free . . . . > The only way to compare the kernels, is to install the standard, enterprise and the 4GB kernel on the same system, and look at it's memoryusage... and you should use the same routines on every kernel testboot, start the same amount of programs, use it the same way, and check the memory usage at the exact same uptime one every kernel..., then you will get the closest match to compare it on... > > You have lots of questions, Robert:) > > > There is no better way to learn, If you don't ask - you'll never know! > I second that ;-) Regards Thomas
