At the time we did the rpm installation we encountered on the /xyz file system transfer rates of about 600-700 KB/sec, which is very poor.

The installation was very slow, especially when using the specific file system.
One of the last steps in the installation is a reboot of the machine, this is to check if the services start in the proper order.
The reboot took a very long time, but after that the server reacted much better.
The observed transfer rates are now in the order of magnitude of 50 MB/sec. Not super, but acceptable.

Although, the problem at this moment no longer exist, I would like to know what could have caused it.
The expected I/O load on the specific file system is high and its performance will directly influence the user experience.
I’d like to prevent the situation that we had during the installation.

Could you please help me on this how to check what caused this issue.

Thanks,
Prem

This could be due to various reasons... aged disk, fragmentation, etc.

Unfortunately the /xyz system is unknown to any of us but yourself; it could
be an nfs mount on a 10MBit link; it could be a USB device ... we wouldn't know.

More detail => better responses.



Cheers,
Tink

yes, I even thought like that but after reboot it seems fine and how to confirm the correct cause of this issue.. Thanks.

Hi Tink,

Its a local filesystem /app/data with 20GB size

For starters:

• What type of filesystem (ext2, xfs, btrfs, fat)?
• What type of device (USB drive, SATA drive), and what brand / RPM is it advertised as?
• What OS / version?
• How are you benchmarking exactly?

•What type of filesystem (ext2, xfs, btrfs, fat)?
ext3
•What type of device (USB drive, SATA drive), and what brand / RPM is it advertised as?
SCSI disks
•What OS / version?
RHEL 5
•How are you benchmarking exactly?
sorry we installing through yum tool and will access the /app/data filesystem for file copies and it noticed.

I've never seen yum echo out disk write speeds ...

# yum install postgresql.x86_64
Resolving Dependencies
Install 2 Package(s)
Is this ok [y/N]: y

Package(s) data still to download: 3.0 M
(1/2): postgresql-9.0.4-5.fc15.x86_64.rpm | 2.8 MB 00:11
(2/2): postgresql-libs-9.0.4-5.fc15.x86_64.rpm | 203 kB 00:00
------------------------------------------------------------------
Total 241 kB/s | 3.0 MB 00:12

Running Transaction
Installing : postgresql-libs-9.0.4-5.fc15.x86_64 1/2
Installing : postgresql-9.0.4-5.fc15.x86_64 2/2


Mentioned Package is just for an example...

Yah

that's your DOWNLOAD speed from a server, not your write-speed to disk.



Cheers,
Tink

Like the man said ...

And it got faster when you re-ran the code because it had already cached the files locally.

Also, this qn
may tell us something

Is /app/data a whole disk or just one partition on a multi-partition disk. ? If the latter, do the other partitions show same speed?

You may want to use this tool to get more info http://linux.die.net/man/8/smartctl