On Thu, 20 Jan 2000, Ray Olszewski wrote about, Re: Detecting Zombies running on the
system?:
> Richard:
>
> You are correct that "free -o" and "top" give the same output. But this form
> of memory-usage reporting is not the best one to see (especially for a
> beginner) -- it gives the impression that the system has less free memory
> than it actually has. The reason is the treatment of cache and buffers.
Well we seem to be in complete agreement now that you say this.
However if one reads the manual, (who does anyway) then it does in certain
terms explane, its just that one (like in all man pages) has to jump into
the skin of the programmer who wrote the dam page).
> As we know, Linux systems increase their efficiency by keeping the code and
> data for recently-run processes in memory. This can cause substantial speed
> improvements -- the quickest way to see this is to mount a new CD, do a
> "find" on it, then do the same "find" again -- the second one will be all
> but instantaneous, while the first one will take some econds to run.
Of course.
>
> But meminfo marks the memory that holds these processes as in use, even
> though it is fully available to new processes that need to run. For a system
> that has been running a long time, the amount of memory "used" in this way
> can be substantial. My own main system, now running for 31 days, shows only
> 4 megs "free" memory (out of 128 megs). But if buffers and cache are taken
> out, the amount of memory actually available jumps to 64 megs.
Yes and no, i understand what you are saying, but a while ago there was
rather a long disscusstion on linux kernel about how this should be
interpreted, however the arguments for and against changing were 50/50
one 50% saying what does it matter, and the other saying you should know
that anyway. (Thats thier optinion) ;-)
>
> My complaint about top (and "free -o") is that is doesn't do the work for
> you of computing this second number, the actual amount of memory available
> for use. You can compute it yourself, from other numbers that are displayed
> ... but why should you? This sort of arithmetic is (part of) what computers
> aare supposed to do for you -- as the output of "free" (without any flags) does.
Well i am lead to belive that one should count the buffers as X amount of
memory, 0ne page, and add it to free memory, then like you say the total
amount of available memory looks far greater.
I rather think its not an issue to a newcommer about how his memory is
being used, but questions do arise from newcommers to linux about what does
it all mean free + cache(available).
I rather think we should not get in too deep here as it is even confusing
to us all sometimes.
>
> At 07:28 AM 1/20/00 +0000, Richard Adams wrote:
> >> Actually, quite a bit less than "all". In particular, top doesn't report the
> >> true amount of available RAM, that is, the amount available including cache
> >> and buffers. Its reporting is equivalent to only the first of the two lines
> >> that "free" reports. Those numbers are VERY misleading, especially on a
> >> system that goeas a long time between reboots.
> >>
> >
> >I dont quite understand why you say this, top gets its info from the same
> >file as free itself which is /proc/meminfo, now i belive top shows memory
> >in the same way as free -o.
> >
> >The buffers line is not shown, but then that information is given per
> >program in tops output.
> >
> >As to the output being different, i dont see your point, if what you say is
> >true, then you must be saying /proc/meminfo does not work properly after x
> >days of uptime.
>
> ------------------------------------"Never tell me the odds!"---
> Ray Olszewski -- Han Solo
> Palo Alto, CA [EMAIL PROTECTED]
> ----------------------------------------------------------------
--
Regards Richard
[EMAIL PROTECTED]
http://people.zeelandnet.nl/pa3gcu/
