Increasing BUFPAGES & NBUF will help somewhat - check the dmesg
	output for a line of the form
	'using 510 buffers containing 2088960 bytes of memory' and set
	them to what you think the first number should be.

	You can also mount the news spool asynchronously (-o async)
	which should really boost the performance of creating/deleting
	files, but at the expense of greater risk of partition
	corruption if the machine crashes.

		David/abs	david@{mono.org,southern.com,mhm-internet.com}

     Microsoft: Asks you where you want to go. NetBSD gets you there.
System Manager: Southern Studios Ltd, PO Box 59, London N22 1AR.
  System Admin: MHM Internet, 14 Barley Mow Passage, Chiswick, London W4 4PH.
         SysOP: Monochrome, Largest UK Internet BBS - 'telnet mono.org'.

On Sun, 2 Mar 1997, John D. Smerdon wrote:

> I am running NetBSD/i386 1.2 on a Pentium 90, 32MB, 2940, and two 4GB
> disks.  The news software is inn 1.5.1 with crosspost, two feeds, 10-20
> readers, and a 10MB history.pag.  The disks are stripped.  The system is
> processing two articles per second, and doing an awful lot of I/O to do it.
>  The system does not receive any binary articles, so the average article
> size is about 4K.  From the stats below it look like it is transferring
> 600K and 80 I/O's per second to process two articles per second.  Does this
> seem normal?  Is there anything I can do to reduce disk activity?
> 
> I will be upgrading the memory to 80MB in the next couple weeks and I
> understand that I need to update NKPDE to 31 in i386/include/pmap.h and set
> 'options EXTMEM_SIZE=80896'  in the config file.  Are there any other
> changes that would improve disk caching?  (NBUF and/or BUFPAGES?)
> 
> Thanks for any assistance.
> 
> Various changes and some 60 second samples:
> 
> kern.maxvnodes = 3000  (Default was around 700)
> 
> Kernel Options:
> options         NMBCLUSTERS=2048
> options         AHC_TAGENABLE
> 
> Filesystem  1K-blocks     Used    Avail Capacity iused   ifree  %iused
> Mounted on
> /dev/sd0a       32399    18405    12374    60%     735    7199     9%   /
> /dev/sd1a       32399    18442    12337    60%     736    7198     9%
> /altroot
> /dev/ccd0d    1556494   824806   653863    56%   52181  324137    14%   /usr
> /dev/ccd1d    6146362  4678564  1160479    80% 1090134  919848    54%   /news
> mfs:21         123887        1   117691     0%       1   30717     0%   /tmp
> 
> /etc/ccd.conf
> ccd0            110     0       /dev/sd0e /dev/sd1e
> ccd1            2090    0       /dev/sd0f /dev/sd1f
> 
> vmstat -w 60 sd0 sd1 ccd0 ccd1
>  procs   memory     page                    disks         faults   cpu
>  r b w   avm   fre  flt  re  pi  po  fr  sr s0 s1 c0 c1   in   sy  cs us sy id
>  0 1 0 72976  4200   68  12   3   1   0   2 52 39  3 86    0  544 110  4 11 85
>  0 1 0 63184  4196    2   0   0   0   0   0 45 40  0 84    0  140  79  3  4 93
>  0 3 0 68768  4196    2   0   0   0   0   0 43 41  0 83    0  156  81  3  4 92
>  0 1 0 73424  4196    2   0   0   0   0   0 46 40  0 84    0  185  88  4  5 91
>  0 0 0 63184  4196    2   0   0   0   0   0 41 39  0 79    0  172  81  4  4 92
>  0 1 0 67840  4196  190  39   9   0   0   0 42 36  1 76    0  209  85  4  8 88
>  0 1 0 73424  4196    1   0   0   0   0   0 34 30  0 64    0  211  86  6  4 90
> 
> systat -w 60 vmstat
>     2 users    Load  1.80  1.49  1.52                  Sun Mar  2 10:55
>  
> Mem:KB  REAL        VIRTUAL                 PAGING   SWAPPING      Interrupts
>       Tot Share    Tot  Share  Free         in  out   in  out           total
> Act 15832   252  82652  43088  4536 count   14
> All 25380  5092 484268 286192       pages
>  
> Proc:r  p  d  s  w    Csw  Trp  Sys  Int  Sof  Flt     55 cow
>               9        99  293  258        49  273     14 objlk
>                                                        11 objht
>    9.9% Sys   4.6% User   0.0% Nice  85.5% Idle        49 zfod
> |    |    |    |    |    |    |    |    |    |    |  5095 nzfod
> =====>>                                              0.96 %zfod
>                                                           kern
> Namei         Sys-cache     Proc-cache               4668 wire
>     Calls     hits    %     hits     %              18028 act
>      4148     3435   83      208     5               2248 inact
>                                                      4536 free
> Discs  fd0  sd0  sd1  ccd  ccd                            daefr
> seeks                                                  84 prcfr
> xfers        45   40    2   81                         52 react
> Kbyte       310  301   12  594                            scan
>   sec       0.6  0.5  0.1  0.9                            hdrev
>                                                           intrn
>  
> 
> 
> --
> John D. Smerdon;  Livonia, Michigan, USA;  Contents are my opinion.
> Home: jds@smerdon.livonia.mi.us
>