Skip to main content

We just started testing our programs on Centos 7 and are finding indexed file reads to be much much slower than from Ubuntu 12.04.  On our original server and dev server, we would handle about 35,000 records in about a second, and now we only get 1000 records a second (just rough timing estimate).   Other programs reading SQL data and file writing and reading thousands of records are very fast, so its just the acucobol thats going slow.  This cobol program is only reading a local file, no databases.  Wondering if anyone has an idea on the issue?

One note, which may be the issue.  For our upgrade we simply copied our runtime to the new server for testing, and looks like verything is working, except performance is bad.  We have version 9.0.1.

We just started testing our programs on Centos 7 and are finding indexed file reads to be much much slower than from Ubuntu 12.04.  On our original server and dev server, we would handle about 35,000 records in about a second, and now we only get 1000 records a second (just rough timing estimate).   Other programs reading SQL data and file writing and reading thousands of records are very fast, so its just the acucobol thats going slow.  This cobol program is only reading a local file, no databases.  Wondering if anyone has an idea on the issue?

One note, which may be the issue.  For our upgrade we simply copied our runtime to the new server for testing, and looks like verything is working, except performance is bad.  We have version 9.0.1.

I would check what file system is being used between the two distro, I think ubuntu uses ext4 and centos uses xfs... this might be a factor to consider.

We just started testing our programs on Centos 7 and are finding indexed file reads to be much much slower than from Ubuntu 12.04.  On our original server and dev server, we would handle about 35,000 records in about a second, and now we only get 1000 records a second (just rough timing estimate).   Other programs reading SQL data and file writing and reading thousands of records are very fast, so its just the acucobol thats going slow.  This cobol program is only reading a local file, no databases.  Wondering if anyone has an idea on the issue?

One note, which may be the issue.  For our upgrade we simply copied our runtime to the new server for testing, and looks like verything is working, except performance is bad.  We have version 9.0.1.

Thanks for the reply. You beat me to it. I solved it , it was filesystem settings. We use ZFS on all the systems, and in this case a its on an SSD that reports 512 byte sectors. Setting ashift=12, for 4K sectors, on pool creation made it work faster, even though the disk reports 512. I ran into a post somewhere about ZFS tips and it mentions how many SSD's need ashift=12 set because internal they are tuned for 4k. I also disabled prefetch. I dont really get why only the cobol files were impacted though, but during a strace I could see it performs pread syscalls, which was different than on other command I was testing. Anyway, this problem is resolved. Thanks again.

We just started testing our programs on Centos 7 and are finding indexed file reads to be much much slower than from Ubuntu 12.04.  On our original server and dev server, we would handle about 35,000 records in about a second, and now we only get 1000 records a second (just rough timing estimate).   Other programs reading SQL data and file writing and reading thousands of records are very fast, so its just the acucobol thats going slow.  This cobol program is only reading a local file, no databases.  Wondering if anyone has an idea on the issue?

One note, which may be the issue.  For our upgrade we simply copied our runtime to the new server for testing, and looks like verything is working, except performance is bad.  We have version 9.0.1.

Great to here you have resolved it and thanks for telling us how you resolved it too.

thanks

Terms & ConditionsAccessibility statement