>On 15 Apr 1996, really kuznet@ms2.inr.ac.ru wrote:
>>
>> Linux really has POOR swapping code, and kswap improved
>> it insignificantly. This opinion is result of every day experience
>> rather than benchmarks. I have Pentium 133 64Mb AIC7881 and run program
>> that eats >=128Mb of virtual memory (certainly, matrices).
>> It sweaps through these matrices with period ~20min.
>Agreed. This is one case where linux doesn't do anything clever at all,
>and one place where we should probably do some work. It happens for a lot
>of scientific calculations.
>The linux swapping code is really written for a more "interactive" load,
>with that kind of access patterns. For that kind of load we do ok, it's
>only a "special case" that we break on (admittedly it's a not-too-unusual
>special case and one we should know about).
[ ... ]
>We should probably modify the LRU algorithm to take RSS into account (and
>possibly access patterns, although that is harder to do). However, to be
>quite honest, my personal priority is not that kind of behaviour, so it
>would be up to somebody else to implement this.. (hint hint)
Would something like madvise be appropriate here?
The reason I ask is that I'm desperately looking for a Linux
VM-related project for a graduate operating systems class. I need to
implement something in about a month (it's taken this long to get
a machine from the department...sigh.)
I'm open to suggestions for anything related to VM, since that's the
code I've studied the most. Any suggestions would be appreciated,
especially if it would lead to something of general interest (i.e. get
put back into the real kernel sometime).
Gideon Glass