Long Checkpoints, balancing network traffic accross multiple Aliases

I have a few questions that I hope someone can answer. Background Info: IDS 7.31FC2 HP/UX 11.0 64 bit 6 CPUs, 5 given to Informix w/ affinity KAIO On OLTP/DSS Mix, configured for OLTP Recently I changed 3 onconfig parameters 19:42:18 Onconfig parameter BUFFERS modified from 200000 to 238400. 19:42:18 Onconfig parameter NUMAIOVPS modified from 1 to 2. 19:42:18 Onconfig parameter SHMADD modified from 153600 to 76800. Before this change the checkpoints were 0 to 1 seconds, now they have jumped to 2 to 5 seconds (with a 25 second checkpoint in there just to spite me). I have cleaners and LRU's maxed to 127 and LRU_MAX set to 1 and LRU_MIN set to 0. Could the addition of 38400 Buffers make this much of a change in checkpoint time? Before Informix would have to write 2000 Buffers at a checkpoint, now it has to write 2286 Buffers, I am having a hard time believing that the additional 286 Buffer writes at checkpoint time are causing a 2 to 3 second checkpoint increase. onstat -p, onstat -m, and the onconfig params follow. Also, I was wondering if it would be helpful to overall performance to configure connections to connect via different ports through multiple aliases. Currently we have 7 machines that connect to our main server, they currently connect to the main server via soctcp through port 1526 server name n2bill1. Would it be helpful to setup /etc/services to contain ports 1526, 1527, 1528, 1529, 1530, 1531, 1532 for soctcp connections to n2bill1 and have those connections map to different DBSERVERALIASES in the sqlhosts file? i.e.. 1526 -> n2bill1_1, 1527 -> n2bill1_2, 1528 -> 1529, etc.? Thanks in advance for any help, Andrew onstat -p Informix Dynamic Server Version 7.31.FC2 -- On-Line -- Up 16:03:44 -- 729576 Kbytes Profile dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached 9998797 15941374 278627430 96.41 566427 989045 1790101 68.36 isamtot open start read write rewrite delete commit rollbk 253258572 428053 2501072 213369175 167719 145009 90428 128 0 gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs 0 0 0 0 0 0 0 ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes 0 0 739 4465.31 1238.77 5 10 bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans 1223950 334 19767729 0 0 121 49793 2880 ixda-RA idx-RA da-RA RA-pgsused lchwaits 7096142 5352 35373 7135187 465107 onstat -m Message Log File: /usr/informix/online.log 09:13:24 Checkpoint Completed: duration was 2 seconds. 09:23:32 Checkpoint Completed: duration was 3 seconds. 09:33:42 Checkpoint Completed: duration was 4 seconds. 09:43:51 Checkpoint Completed: duration was 2 seconds. 09:54:00 Checkpoint Completed: duration was 2 seconds. 10:04:09 Checkpoint Completed: duration was 3 seconds. 10:14:19 Checkpoint Completed: duration was 4 seconds. 10:24:27 Checkpoint Completed: duration was 2 seconds. 10:34:35 Checkpoint Completed: duration was 2 seconds. 10:44:44 Checkpoint Completed: duration was 2 seconds. 10:54:53 Checkpoint Completed: duration was 3 seconds. 11:05:02 Checkpoint Completed: duration was 3 seconds. 11:13:55 listener-thread: err = -951: oserr = 0: errstr = Admin: Incorrectpassword or user Admin is not known on the database server. 11:13:56 listener-thread: err = -951: oserr = 0: errstr = Admin: Incorrectpassword or user Admin is not known on the database server. 11:15:10 Checkpoint Completed: duration was 2 seconds. 11:25:20 Checkpoint Completed: duration was 2 seconds. 11:35:28 Checkpoint Completed: duration was 2 seconds. 11:45:38 Checkpoint Completed: duration was 2 seconds. Onconfig Params ROOTNAME rootdbsROOTPATH /dev/informix_space/rootchunk DBSERVERNAME n2bill1_shm MIRRORPATH \\PHYSDBS bigphys MSGPATH /usr/informix/online.log CONSOLE /dev/console TAPEDEV /dev/null LTAPEDEV /dev/null ROOTOFFSET 0 ROOTSIZE 100000MIRROR 0 MIRROROFFSET 0 PHYSFILE 159894 LOGFILES 32 LOGSIZE 1500 TAPEBLK 16 TAPESIZE 10240 LTAPEBLK 16 LTAPESIZE 10240 SERVERNUM 0 DEADLOCK_TIMEOUT 60 RESIDENT 1 LOCKS 1048576 BUFFERS 238400 ONDBSPACEDOWN 2LBU_PRESERVE 1 OPCACHEMAX 0 PHYSBUFF 32 LOGBUFF 32LOGSMAX 50 CLEANERS 127 SHMBASE 0xc000000000232000 CKPTINTVL 600 LRUS 127 LRU_MAX_DIRTY 1 LRU_MIN_DIRTY 0 SINGLE_CPU_VP 0 LTXHWM 50 LTXEHWM 60 TXTIMEOUT 300 NUMCPUVPS 5 DRAUTO 0 SHMVIRTSIZE 128000 USEOSTIME 0 NOAGE 1 AFF_SPROC 1 AFF_NPROCS 5 RA_PAGES 64 RA_THRESHOLD 42 NUMAIOVPS 2 NETTYPE ipcshm,,50,CPU NETTYPE soctcp,,25,NET DBSERVERALIASES n2bill1 MULTIPROCESSOR 1 FILLFACTOR 80 STACKSIZE 64DBSPACETEMP tempdbs1,tempdbs2,tempdbs3,tempdbs4 STAGEBLOB \\ DRNODE 1 DRNAME \\ DRINTERVAL 30 DRTIMEOUT 30DRLOSTFOUND /usr/informix/etc/dr.lostfound OFF_RECVRY_THREADS 50 ON_RECVRY_THREADS 5 DUMPSHMEM 0 DUMPGCORE 0 DUMPCORE 0 DUMPCNT 1DUMPDIR /var/simplified/log/dump DATASKIP off DS_MAX_QUERIES 1024 DS_TOTAL_MEMORY 131072 DS_MAX_SCANS 100000 SHMADD 0x4b00000 SHMTOTAL 0x0 OPTCOMPIND 2 MAX_PDQPRIORITY 100 HETERO_COMMIT 0 BAR_ACT_LOG /tmp/bar_act.log BAR_DEBUG_LOG /tmp/bar_dbug.log BAR_MAX_BACKUP 0B

Andrew Ford wrote: > > I have a few questions that I hope someone can answer. > > Background Info: > > IDS 7.31FC2 > HP/UX 11.0 64 bit > 6 CPUs, 5 given to Informix w/ affinity > KAIO On > OLTP/DSS Mix, configured for OLTP > > Recently I changed 3 onconfig parameters > 19:42:18 Onconfig parameter BUFFERS modified from 200000 to 238400. > 19:42:18 Onconfig parameter NUMAIOVPS modified from 1 to 2. > 19:42:18 Onconfig parameter SHMADD modified from 153600 to 76800.> > Before this change the checkpoints were 0 to 1 seconds, now they have > jumped to 2 to 5 seconds (with a 25 second checkpoint in there just to > spite me). I have cleaners and LRU's maxed to 127 and LRU_MAX set to 1 and > LRU_MIN set to 0. You have 7.31 so you can go for 128 without triggering that old annoying 0 length Physical Log bug, FWIW. > Could the addition of 38400 Buffers make this much of a change in > checkpoint time? Before Informix would have to write 2000 Buffers at a > checkpoint, now it has to write 2286 Buffers, I am having a hard time > believing that the additional 286 Buffer writes at checkpoint time are > causing a 2 to 3 second checkpoint increase. Does seem odd. > onstat -p, onstat -m, and the onconfig params follow.> > Also, I was wondering if it would be helpful to overall performance to > configure connections to connect via different ports through multiple > aliases. > > Currently we have 7 machines that connect to our main server, they > currently connect to the main server via soctcp through port 1526 server > name n2bill1. > Would it be helpful to setup /etc/services to contain ports 1526, 1527, > 1528, 1529, 1530, 1531, 1532 for soctcp connections to n2bill1 and have > those connections map to different DBSERVERALIASES in the sqlhosts file? > i.e.. 1526 -> n2bill1_1, 1527 -> n2bill1_2, 1528 -> 1529, etc.? Only if you have multiple interfaces on the machine and there seems to be a network bottleneck. > Thanks in advance for any help, Let's see onstat -F and onstat -R (from shortly before a scheduled checkpoint not just after), also onstat -g glo and onstat -g iov. Art S. Kagel