Kernel:                 2.4.15.pre7
Total mem:              9GB
Total Swap:             5GB
Total CPU:              2 (Pentium III coppermine)
Total exec time:        24Hrs

I executed my memory management tests on linux-2.4.15.pre7.

The tests can be found on ltp web site, ltp.sf.net, under ../testcases/kernel/
mem/... and .../testcases/kernel/network/nfs/nfsstress/...

The programs do the following:

   1) mmap a file of size = sizeof(char) * 1000000000; as shared and private
   mapping. (anonymous memory map)

   3) mmap a file of same size shared.

   4) memory race conditions tests.

   5) spawn 30 threads, each thread mallocs memory in a loop size is either
   powers of 2,3,or 7 or s
   6) Also I am generating a lot of disk I/O and compiles (invoking cc) by
   running my make_tree (NFS client) test over NFS. (details about the test
   available in the source .../testcases/kernel/network/nfs/nfsstress/
   make_tree.c.)  

The memory tests are executed as separate processes as a normal user "manjo".
The nfs test is executed as a root process on nfs version 2.

The tests were set to execute for a period of 24 hours. I noticed that the
network was taken down, my telnet sessions that were displaying out put of
top command hung. The ethernet driver was spitting messages to the console;

eth0: card reports no resources.

All but one of my tests hung. Test 5 was running at the end of 24hrs. I tried
to kill my test but even then the machine did not recover from the state it was
in. The response from the machine was very very slow. When I tried to reboot
the machine, after the message "system is now going down ...." The was no
activity. At regular interavels I was getting messages from the ethernet driver

eth0: card reports no resources.

Also, I was getting messages from the NFS server:
nfs: task 60771 cannot get a request slot.

The same tests on 2.4.15.pre6 did not cause the machine to behave in this manner
The tests ended gracefully, or I was able to kill running tests and go on with
collecting data, on 2.4.15.pre7 I had to hard reboot the machine inorder to
gather data that was logged to a file using sar.
The results for 2.4.15.pre6 can be found in the ltp
mailing list at ltp.sf.net

I am including below various statics I gathered using sar.

I/O  and transfer rate statistics
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25          tps      rtps      wtps   bread/s   bwrtn/s
Average:       115.10     70.42     44.68   4217.46  15697.65

paging statistics
~~~~~~~~~~~~~~~~~
14:01:25     pgpgin/s pgpgout/s  activepg  inadtypg  inaclnpg  inatarpg
Average:      2108.73   7848.74     48797         0         0         0

process creation activity.
~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25       proc/s
Average:         8.60


activity for each block device
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25       DEV       tps     blks/s
Average:       dev8-0    115.10  19915.10

network statistics.
~~~~~~~~~~~~~~~~~~~
DEV
~~~
14:01:25        IFACE   rxpck/s   txpck/s   rxbyt/s   txbyt/s   rxcmp/s  txcmp/s
                rxmcst/s
Average:        eth0    187.61    188.82    33267.12  33193.28   0.00     0.00
                0.00

EDEV
~~~~
14:01:25        IFACE   rxerr/s   txerr/s   coll/s  rxdrop/s  txdrop/s  txcarr/s
                rxfram/s  rxfifo/s  txfifo/s
Average:         eth0    0.02      0.00     0.00    0.00       0.00     0.00
                0.00      0.00

SOCK
~~~~
14:01:25       totsck    tcpsck    udpsck    rawsck   ip-frag
Average:           48        17        14         1         0

queue length and load averages
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25      runq-sz  plist-sz   ldavg-1   ldavg-5
Average:           22       442     36.72     36.61

memory  and  swap space utilization statistics
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25    kbmemfree kbmemused  %memused kbmemshrd kbbuffers kbcached kbswpfree
            kbswpused  %swpused
Average:    110209    919147     89.29    0         1250       401494   332314
            205822     38.25

memory statistics.
~~~~~~~~~~~~~~~~~~
14:01:25      frmpg/s   shmpg/s   bufpg/s   campg/s
Average:        -1.08      0.00      0.00      0.14

CPU  utilization.
~~~~~~~~~~~~~~~~~
14:01:25          CPU     %user     %nice   %system     %idle
Average:          all      7.52      0.00     64.45     28.03

verage:            0      7.66      0.00     64.32     28.02
Average:           1      7.38      0.00     64.58     28.04

status  of  inode,  file  and  other kernel tables.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25    dentunusd   file-sz  %file-sz  inode-sz  super-sz %super-sz dquot-sz
            %dquot-sz  rtsig-sz %rtsig-sz
Average:    160         4985     60.85       861         0      0.00         0
            0.00         1      0.00

system switching activity.
~~~~~~~~~~~~~~~~~~~~~~~~~~
14:01:25      cswch/s
Average:      4578.44

swapping  statistics.
~~~~~~~~~~~~~~~~~~~~~
14:01:25     pswpin/s pswpout/s
Average:         3.93    856.40


Thanks
-- Manoj

***********************************************************
        The greatest risk is in not taking one
***********************************************************