Very Large "zero" slab expected?

[Sonny Rao]

On Tue, Feb 01, 2005 at 02:52:07PM -0500, Sonny Rao wrote:
> Hi, I'm running a fairly heavy database workload using DB/2 on a 4-way
> 720 ontop of ext3 trying to replicate a separate VM performance issue.
> 
> During and after the run, I see a massive about of slab memory being
> used by the "zero" cache, which appears to a be a ppc64 specific
> cache for pmd's and pgd's.  
> 
> Basically, about 1GB out of my total of 8GB on the system is being
> consumed by this.  Is this expected ?  It seems excessive
> considering my mapped memory is only about 2GB during the run.
> 
> The DB/2 buffer pool was set to 4GB and none of it used hugh pages.
> 
> Here's the relavent snippet from slabinfo:
> slabinfo - version: 2.0
> # name            <active_objs> <num_objs> <objsize> <objperslab> <pagesperslab> : tunables <batchcount> <limit> <sharedfactor> : slabdata <active_slabs> <num_slabs> <sharedavail>
> zero              256424 256431   4096    1    1 : tunables   24   12 8 : slabdata 256424 256431      0
> 
> Any ideas as to why there are so many of these ?

More info :

Here's /proc/meminfo output from a similar run, with 4GB of RAM
instead of 8GB:  

portrait:~ # cat /proc/meminfo 
MemTotal:      4099072 kB
MemFree:         15008 kB
Buffers:         38580 kB
Cached:        2941320 kB
SwapCached:     118492 kB
Active:        2672776 kB
Inactive:       645284 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:      4099072 kB
LowFree:         15008 kB
SwapTotal:     1050616 kB
SwapFree:       341156 kB
Dirty:             372 kB
Writeback:           4 kB
Mapped:        1827844 kB
Slab:           715384 kB
Committed_AS:  5342648 kB
PageTables:     610864 kB
VmallocTotal: 2147483647 kB
VmallocUsed:     25340 kB
VmallocChunk: 2147458223 kB
HugePages_Total:     0
HugePages_Free:      0
Hugepagesize:    16384 kB

And the top few lines of slabtop:

 Active / Total Objects (% used)    : 728019 / 747302 (97.4%)
 Active / Total Slabs (% used)      : 177511 / 177591 (100.0%)
 Active / Total Caches (% used)     : 79 / 120 (65.8%)
 Active / Total Size (% used)       : 705236.61K / 709101.23K (99.5%)
 Minimum / Average / Maximum Object : 0.02K / 0.95K / 128.00K

  OBJS ACTIVE  USE OBJ SIZE  SLABS OBJ/SLAB CACHE SIZE NAME                   
336360 335400  99%    0.09K   8409       40     33636K buffer_head
157860 156817  99%    0.12K   5262       30     21048K size-128
154036 153988  99%    4.00K 154036        1    616144K zero
 22572  22120  97%    0.16K   1026       22      4104K vm_area_struct
 18571  18074  97%    0.52K   2653        7     10612K radix_tree_node


So, you can see at this point we have around 600MB of these "zero"
pages being used.  

There are about 150 db2 processes running using lots of shared memory: 

portrait:~ # ipcs 

------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch
status      
0x763ee574 2392064    db2inst1  767        39987136   161                     
0x763ee561 2424833    db2inst1  701        39370752   159                     
0x00000000 2457602    db2fenc1  701        164052992  161                     
0x6803bbdb 2490371    db2inst1  761        50331648   1                       
0x00000000 2785285    db2inst1  701        1073741824 145                     
0x00000000 2818054    db2inst1  701        1073741824 145                     
0x00000000 2850823    db2inst1  701        1073741824 145                     
0x00000000 2883592    db2inst1  701        117735424  145    


Thanks,

Sonny