IDS Forum

Re: IO performance issues

Posted By: Art Kagel
Date: Thursday, 8 December 2016, at 6:44 a.m.

In Response To: IO performance issues (R MICHAEL)

Michael:

Are these really your BUFFERPOOL settings?

BUFFERPOOL default,buffers=10000,lrus=8,lru_min_dirty=50,lru_ max_dirty=60.5
BUFFERPOOL size=2K
BUFFERPOOL size=8K

IB that this implies that you have 10,000 2K and 10,000 8K buffers. The
write cache % of 69% indicates that you likely need many more buffers.
However, there's not enough information here to determine how many nor to
which buffer pool they need to be added if not both. Also, I see no
indication that there is an IO problem, but you didn't post the onstat
output we would need to determine that, nor for that matter those we'd need
to determine what else (besides not enough buffer cache) is your problem.

I think what you mean to say is that you have a performance problem that
MIGHT be IO related, but you don't know. Honestly, and I'm not just angling
for a paid gig here, you need someone like me or Lester or one of the other
excellent consultants who specialize in performance tuning to do a full on
server health check for you.

That said, zero your stats (onstat -z) at the end of today's normal
processing cycle and then at the end of the next day capture and post the
following output (and include the onstat header lines please - there's info
there that we need) and the number of hours between zeroing the stats and
taking those readings for us and we'll see if anyone in the community can
spot the problem:

onstat -p (again)
onstat -g buf
onstat -g iov
onstat -g iof
onstat -D
onstat -g seg
onstat -g glo
onstat -g rea
onstat -g ckp (during normal processing hours)'
onstat -g dic
onstat -g dsc
onstat -g prc

Art

Art S. Kagel, President and Principal Consultant
ASK Database Management
www.askdbmgt.com

Blog: http://informix-myview.blogspot.com/

Disclaimer: Please keep in mind that my own opinions are my own opinions
and do not reflect on the IIUG, nor any other organization with which I am
associated either explicitly, implicitly, or by inference. Neither do
those opinions reflect those of other individuals affiliated with any
entity with which I am affiliated nor those of the entities themselves.

On Thu, Dec 8, 2016 at 5:47 AM, R MICHAEL <mboncalo@gmail.com> wrote:

> I'm opening this thread because of some IO performance issues that I have
> now
> and need to be sure of what I must do.
>
> This is my db profile counts:Profile
> dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached
> 22906487 87848814 34363980245 99.93 332057743 398328761 1082072410 69.31
>
> isamtot open start read write rewrite delete commit rollbk
> 48406998458 7423608018 9188986012 10875359326 1488886 347373130 682316
> 347179506 1
>
> gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs
> 1487097310 353965214 343076548 4397605 4723657 2 18018239
>
> ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes
> 0 0 0 457810.60 98673.63 3364 7580
>
> bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans
> 541933 6627626 13923560081 0 0 29422 1170160 807177
>
> ixda-RA idx-RA da-RA logrec-RA RA-pgsused lchwaits
> 70449 9643446 760355 1294 18020622 137838512
>
> Onconfig:
> NETTYPE soctcp,1,200,NET
> LISTEN_TIMEOUT 60
> MAX_INCOMPLETE_CONNECTIONS 1024
> FASTPOLL 1
> NUMFDSERVERS 4
> NS_CACHE host=900,service=900,user=900,group=900
>
> MULTIPROCESSOR 1
> VPCLASS cpu,num=12,noage
> VP_MEMORY_CACHE_KB 0
> SINGLE_CPU_VP 0
> CLEANERS 8
> DIRECT_IO 1
> LOCKS 20000
> DEF_TABLE_LOCKMODE page
> RESIDENT 0
> SHMBASE 0x44000000L
> SHMVIRTSIZE 10485760
> SHMADD 262144
> EXTSHMADD 51200
> SHMTOTAL 62914560
> SHMVIRT_ALLOCSEG 0,3
> SHMNOACCESS
> MAX_PDQPRIORITY 100
> DS_MAX_QUERIES
> DS_TOTAL_MEMORY
> DS_MAX_SCANS 1048576
> DS_NONPDQ_QUERY_MEM 256
> DATASKIP
> BUFFERPOOL default,buffers=10000,lrus=8,lru_min_dirty=50,lru_
> max_dirty=60.5
> BUFFERPOOL size=2K
> BUFFERPOOL size=8K
>
> onstat -P
> Percentages:
> Data 55.67
> Btree 8.09
> Other 36.24
>
> onstat -F
>
> Fg Writes LRU Writes Chunk Writes
> 2 611239 14997845
>
> address flusher state data # LRU Chunk Wakeups Idle Tim
> 4c3e48e8 0 I 0 603 10296 1021956 1010204.637
> 4c3e51a8 1 I 0 707 9448 1021391 1009724.104
> 4c3e5a68 2 I 0 627 7475 1019552 1010959.875
> 4c3e6328 3 I 0 873 4565 1017161 1011482.790
> 4c3e6be8 4 I 0 742 7209 1019094 1010744.211
> 4c3e74a8 5 I 0 727 6145 1018085 1010709.286
> 4c3e7d68 6 I 0 858 7217 1019100 1010542.952
> 4c3e8628 7 I 0 773 6640 1018665 1010826.101
> states: Exit Idle Chunk Lru
>
> The machine on which DB runs has 16 cpus and 64 GB of RAM.
>
> These values are not configured by me.
>
> Any ideeas ?
>
> Thanks.
>
>
> ************************************************************
> *******************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>

--001a1148d7be93048d0543242905

Messages In This Thread

• IO performance issues
  R MICHAEL -- Thursday, 8 December 2016, at 5:47 a.m.
  • Re: IO performance issues
    R MICHAEL -- Thursday, 8 December 2016, at 6:15 a.m.
  • Re: IO performance issues
    Art Kagel -- Thursday, 8 December 2016, at 6:44 a.m.
  • Re: IO performance issues
    Art Kagel -- Thursday, 8 December 2016, at 6:55 a.m.
    • Re: IO performance issues
      R MICHAEL -- Wednesday, 21 December 2016, at 4:14 p.m.
      • Re: IO performance issues
        Marcus Haarmann -- Wednesday, 21 December 2016, at 5:19 p.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.