Regading to io/wup

Hi, I have 10000 buffers, 10 LRUs, 10 Cleaners and initially 10 vps AIO (no Kaio, 1 processor, 128 Ram, SuSE Linux, IDS 7.30.UC7, 10 users) I've read in this group that at least, 1 io/wup should be to < 1.0, so I added more AIO vps. Firstly 2, then 2 plus and so on, until 16 ... but no io/wup is less to 1.0 yet ... so, Must I go on adding more AIOs, or that is not really so important ? (actually the system works quite well ...) Ok, Thanks, Manel Falco Read Utilization (UR) 100% O.K.: 99.8962 % Bufwaits Ratio (BR) < 7% O.K.: 1.09541 % $ onstat -g iov AIO I/O vps: class/vp s io/s totalops dskread dskwrite dskcopy wakeups io/wup errors msc 0 i 0.0 124 0 0 0 124 1.0 0 aio 0 i 0.8 69072 67505 1537 0 59319 1.2 0 aio 1 i 0.7 57532 55864 1668 0 55301 1.0 0 aio 2 i 0.6 51112 49897 1214 0 48778 1.0 0 aio 3 i 0.5 41845 40949 896 0 42207 1.0 0 aio 4 i 0.4 34655 33672 983 0 33986 1.0 0 aio 5 i 0.3 25143 24262 881 0 24713 1.0 0 aio 6 i 0.3 20889 20201 688 0 20491 1.0 0 aio 7 i 0.2 18665 18204 461 0 18465 1.0 0 aio 8 i 0.2 17552 16990 562 0 17229 1.0 0 aio 9 i 0.2 16224 15943 281 0 15870 1.0 0 aio 10 i 0.2 15545 15273 272 0 15213 1.0 0 aio 11 i 0.2 14930 14680 250 0 14591 1.0 0 aio 12 i 0.2 14479 14242 237 0 13993 1.0 0 aio 13 i 0.2 14047 13815 232 0 13696 1.0 0 aio 14 i 0.2 13454 13235 219 0 12940 1.0 0 aio 15 i 0.2 12821 12620 201 0 12457 1.0 0 pio 0 i 0.0 412 0 412 0 412 1.0 0 lio 0 i 0.1 7855 0 7855 0 7855 1.0 0 $ onstat -g ioq AIO I/O queues:q name/id len maxlen totalops dskread dskwrite dskcopy adt 0 0 0 0 0 0 0 msc 0 0 1 128 0 0 0 aio 0 0 1 64 32 0 0 pio 0 0 1 424 0 424 0 lio 0 0 1 8475 0 8475 0 gfd 3 0 89 1395 450 945 0 gfd 4 0 0 0 0 0 0 gfd 5 0 167 751870 743427 8443 0 gfd 6 0 129 7137 3032 4105 0

To handle peak loads it looks like you still need more AIO VPs, yes. BTW that chunk#5 is being hit awfully hard, consider shifting some tables around or fragmenting or getting another drive, something. Art S. Kagel "Manel Falcó i Aige" wrote: > > Hi, > I have 10000 buffers, 10 LRUs, 10 Cleaners and initially 10 vps AIO > (no Kaio, 1 processor, 128 Ram, SuSE Linux, IDS 7.30.UC7, 10 users) > > I've read in this group that at least, 1 io/wup should be to < 1.0, > so I added more AIO vps. Firstly 2, then 2 plus and so on, until 16 > ... but no io/wup is less to 1.0 yet ... so, Must I go on adding more > AIOs, or that is not really so important ? (actually the system works > quite well ...) > > Ok, Thanks, > Manel Falco > > Read Utilization (UR) 100% O.K.: 99.8962 % > Bufwaits Ratio (BR) < 7% O.K.: 1.09541 % > > $ onstat -g iov > AIO I/O vps:> > class/vp s io/s totalops dskread dskwrite dskcopy wakeups io/wup > errors > msc 0 i 0.0 124 0 0 0 124 1.0 > 0 > aio 0 i 0.8 69072 67505 1537 0 59319 1.2 > 0 > aio 1 i 0.7 57532 55864 1668 0 55301 1.0 > 0 > aio 2 i 0.6 51112 49897 1214 0 48778 1.0 > 0 > aio 3 i 0.5 41845 40949 896 0 42207 1.0 > 0 > aio 4 i 0.4 34655 33672 983 0 33986 1.0 > 0 > aio 5 i 0.3 25143 24262 881 0 24713 1.0 > 0 > aio 6 i 0.3 20889 20201 688 0 20491 1.0 > 0 > aio 7 i 0.2 18665 18204 461 0 18465 1.0 > 0 > aio 8 i 0.2 17552 16990 562 0 17229 1.0 > 0 > aio 9 i 0.2 16224 15943 281 0 15870 1.0 > 0 > aio 10 i 0.2 15545 15273 272 0 15213 1.0 > 0 > aio 11 i 0.2 14930 14680 250 0 14591 1.0 > 0 > aio 12 i 0.2 14479 14242 237 0 13993 1.0 > 0 > aio 13 i 0.2 14047 13815 232 0 13696 1.0 > 0 > aio 14 i 0.2 13454 13235 219 0 12940 1.0 > 0 > aio 15 i 0.2 12821 12620 201 0 12457 1.0 > 0 > pio 0 i 0.0 412 0 412 0 412 1.0 > 0 > lio 0 i 0.1 7855 0 7855 0 7855 1.0 > 0 > > $ onstat -g ioq > AIO I/O queues:> q name/id len maxlen totalops dskread dskwrite dskcopy > adt 0 0 0 0 0 0 0 > msc 0 0 1 128 0 0 0 > aio 0 0 1 64 32 0 0 > pio 0 0 1 424 0 424 0 > lio 0 0 1 8475 0 8475 0 > gfd 3 0 89 1395 450 945 0 > gfd 4 0 0 0 0 0 0 > gfd 5 0 167 751870 743427 8443 0 > gfd 6 0 129 7137 3032 4105 0

Where you saw anything about the chunk #5 ?? by onstat -g iov ? -> gfd 5 0 167 751870 743427 8443 0 I know, I know the more machine the better ... ;o) but if not possible, Would it improve if I add more chunks for data ?? BTW 128 Ram -> 10000 buffers -> 10 LRU -> 10 Cleaners OK ? Thanks, Manel Falcó On Fri, 11 Aug 2000 13:47:32 -0400, "Art S. Kagel" <kagel@bloomberg.net> wrote: >To handle peak loads it looks like you still need more AIO VPs, yes. >BTW that chunk#5 is being hit awfully hard, consider shifting some >tables around or fragmenting or getting another drive, something. > >Art S. Kagel > >

"Manel Falcó i Aige" wrote: > > Where you saw anything about the chunk #5 ?? > by onstat -g iov ? -> gfd 5 0 167 751870 743427 8443 0 The onstat -g ioq you posted. Also look at onstat -D & onstat -g iof. > I know, I know the more machine the better ... ;o) > but if not possible, Would it improve if I add more chunks for data ?? Only if they are on separate spindles. > BTW 128 Ram -> 10000 buffers -> 10 LRU -> 10 Cleaners OK ? Unfortunately you will only know if it is OK if you don't have any performance problems. Art S. Kagel > Thanks, > Manel Falcó > > On Fri, 11 Aug 2000 13:47:32 -0400, "Art S. Kagel" > <kagel@bloomberg.net> wrote: > > >To handle peak loads it looks like you still need more AIO VPs, yes. > >BTW that chunk#5 is being hit awfully hard, consider shifting some > >tables around or fragmenting or getting another drive, something. > > > >Art S. Kagel > > > >

Art S. Kagel wrote in message >> BTW 128 Ram -> 10000 buffers -> 10 LRU -> 10 Cleaners OK ? > >Unfortunately you will only know if it is OK if you don't have any >performance problems. > >Art S. Kagel > What type of machine is this? I would normally go 127 LRUS 48 CLEANERS What are the checkpoint times?

> > What type of machine is this? I would normally go > 127 LRUS 48 CLEANERS > What are the checkpoint times? My Box: 1 processor 450, 128 Ram, SuSE Linux 6.4 (No KAIO), IDS 7.30.UC7, 10,15 users, 10000 buffers, 10 LRU/Cleaner, CKPTINTVL=300 LRU_MAX=2, MIN=1. Checkpoints of 0,1 to 3 seconds. Ok. I increased up to 20 vps aio !! and the most are to 1.0 but none to less ...therefore there are 20 aio Unix process ... so, must I add more ? more overhead then ? too many Unix process ? Why you recommend so many LRU / Cleaners ? more overhead ? Regards, Manel Falcó On Mon, 14 Aug 2000 22:44:27 +0100, "smooth1" <smooth1@iclway.co.uk> wrote: > >Art S. Kagel wrote in message >> BTW 128 Ram -> 10000 buffers -> 10 LRU -> >10 Cleaners OK ? >> >>Unfortunately you will only know if it is OK if you don't have any >>performance problems. >> >>Art S. Kagel >> > > What type of machine is this? I would normally go > 127 LRUS 48 CLEANERS > What are the checkpoint times? > >

Because it improves server performance. What are you saving the cycles for? On a dedicated server there should not be anything else running. Art S. Kagel "Manel Falcó i Aige" wrote: > > > > > What type of machine is this? I would normally go > > 127 LRUS 48 CLEANERS > > What are the checkpoint times? > My Box: > 1 processor 450, 128 Ram, SuSE Linux 6.4 (No KAIO), IDS 7.30.UC7, > 10,15 users, 10000 buffers, 10 LRU/Cleaner, CKPTINTVL=300 LRU_MAX=2, > MIN=1. Checkpoints of 0,1 to 3 seconds. Ok. > > I increased up to 20 vps aio !! and the most are to 1.0 but none to > less ...therefore there are 20 aio Unix process ... so, must I add > more ? more overhead then ? too many Unix process ? > > Why you recommend so many LRU / Cleaners ? more overhead ? > > Regards, > Manel Falcó > > On Mon, 14 Aug 2000 22:44:27 +0100, "smooth1" <smooth1@iclway.co.uk> > wrote: > > > > >Art S. Kagel wrote in message >> BTW 128 Ram -> 10000 buffers -> 10 LRU -> > >10 Cleaners OK ? > >> > >>Unfortunately you will only know if it is OK if you don't have any > >>performance problems. > >> > >>Art S. Kagel > >> > > > > What type of machine is this? I would normally go > > 127 LRUS 48 CLEANERS > > What are the checkpoint times? > > > >