RE: Virtual Memory
- From: "Kevin Closson" <kevinc@xxxxxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Fri, 25 Aug 2006 21:00:46 -0700
>>Interestingly enough, in the "allocate-swap-on-demand" camp,
the most common "algorithm" I recall for deciding which process to kill
was quite simple. In order to mimimise the number of processes to be
killed, the OS would always kill the one with the largest memory
footprint first. On a database server, you can probably guess what this
will be...
...all the allocate-on-demand Unix derivations that I had first hand
experience with
selected processes with the largest resident set (minus IPC
shared memory and mmap pages), and the most sleep. Database server
processes always
got whacked. I worked for a systems house that made servers
SPECIFICALLY for
database and most optimized for Oracle--Sequent. We, uh, preallocated
swap since
to us there was no such thing as a run of the mill fat, sleepy
process...since that
was generally an Oracle process (and background at that) anyway when you
look closely ...
...ahhh, memory lane..walking down memory lane would be strong,
effective medicine for
the folks that are shoveling code into the Linux kernel....
...Linux does not preallocate swap (within the page allocator) :-(
--
//www.freelists.org/webpage/oracle-l
Other related posts:
