checkpoint/load/physical log question

hi all: I am trying to load about 1 gig data into 8 different tables from a directory. It takes about 8 hours. When I check the cpu usage, I notice most of the cpu is wio (waiting for i/o) about 95% of it, what does this indicate? I also checked my online log, it seems to do a checkpoint entry every 2 minutes when I load, could that slow down the load process? since my checkpoint interval is set to be 30 mins, the only other explanation is that the physical log is over 75% full. how do you exactly find out the fullness of the physical log and how do you reduct that? any other things I should check into? thanks. yan

Yan - Sounds like you're running into disk contention. First off, the only way to control how quickly your physical log fills (without changing _what_ you're doing) is to change its size. Sounds to me like you might seriously consider increasing its size. Next things to look for would be physical locations of the data that you're working with. Are the physical and logical logs on separate physical disks (not logical, _physical_)? If not, this will hurt performance. Also, is your data spread out, or all on one disk? Check to see if you're using KAIO. If you aren't make sure you've got enough AIO VPs configured. If you _are_ using KAIO, make sure that you're actually getting kaio threads. If not, then KAIO isn't happening, regardless of the settings in ONCONFIG (found this out the hard way on Digital platform). Next, you might try running a 'sar -d' to determine which disks are bottlenecked. Hope this helps. Let me know in private e-mail if you have any other questions or want more detail. Yan Zhu wrote in message <77ll0t$gdj$1@news.xmission.com>... > > >hi all: > I am trying to load about 1 gig data into 8 different tables from a >directory. It takes about 8 hours. When I check the cpu usage, I notice >most of the cpu is wio (waiting for i/o) about 95% of it, what does this >indicate? > I also checked my online log, it seems to do a checkpoint entry >every 2 minutes when I load, could that slow down the load process? > since my checkpoint interval is set to be 30 mins, the only other >explanation is that the physical log is over 75% full. >how do you exactly find out the fullness of the physical log and how do >you reduct that? > any other things I should check into? >thanks. >yan > >

Also, the top section of onstat -l will show you the total and used space in pages of your physical log. Thomas J. Girsch wrote in message <369e5a02.0@newsfeed.one.net>... >Yan - > >Sounds like you're running into disk contention. > >First off, the only way to control how quickly your physical log fills >(without changing _what_ you're doing) is to change its size. Sounds to me >like you might seriously consider increasing its size. > >Next things to look for would be physical locations of the data that you're >working with. Are the physical and logical logs on separate physical disks >(not logical, _physical_)? If not, this will hurt performance. Also, is >your data spread out, or all on one disk? > >Check to see if you're using KAIO. If you aren't make sure you've got >enough AIO VPs configured. If you _are_ using KAIO, make sure that you're >actually getting kaio threads. If not, then KAIO isn't happening, >regardless of the settings in ONCONFIG (found this out the hard way on >Digital platform). > >Next, you might try running a 'sar -d' to determine which disks are >bottlenecked. > >Hope this helps. Let me know in private e-mail if you have any other >questions or want more detail. >Yan Zhu wrote in message <77ll0t$gdj$1@news.xmission.com>... >> >> >>hi all: >> I am trying to load about 1 gig data into 8 different tables from a >>directory. It takes about 8 hours. When I check the cpu usage, I notice >>most of the cpu is wio (waiting for i/o) about 95% of it, what does this >>indicate? >> I also checked my online log, it seems to do a checkpoint entry >>every 2 minutes when I load, could that slow down the load process? >> since my checkpoint interval is set to be 30 mins, the only other >>explanation is that the physical log is over 75% full. >>how do you exactly find out the fullness of the physical log and how do >>you reduct that? >> any other things I should check into? >>thanks. >>yan >> >> > >