High I/O Wait when copying/moving Files

Latest response

Hi All,

I am currently experiencing a very high IO wait whenever I copy or move a large file from a directory (say /home/user1) to another directory (say /home/user2). WA percentage in top command reaches more than 50%. This disables and hangs the desktop. I read somewhere that during copying of files, large percentage of my memory is used for cache. Is it possible to set a certain amount of my memory (8 Gb) during copy operations? I know cache memory is automatically given back to the system, but I want to allocate specific amount during copy so the Redhat desktop does not experience hang and unresponsiveness.

Responses

Not really giving a lot of information to go on, here. Just to start, some of the things that might be helpful in providing an answer would be:
* What version of Red Hat
* What patch-set has been applied
* What filesystem type (extN, xfs, etc.)
* What filesystem mount-options are in us
* How are you copying files
* How are you moving files
* Is there a performance difference between copying and moving
* Are the source and destination on separate filesystems
* What's the underlying storage look like for the source/destination
* Do you have A/V enabled on your system (e.g., I've seen McAfee absolutely murder performance on both copy and move operations)

Hi,

Sorry about the lack of information. Please see all the details you requested.

  • What version of Red Hat
  • Red Hat Enterprise Linux 7.0 (Maipo)

  • What patch-set has been applied

  • No patch-set applied yet

  • What filesystem type (extN, xfs, etc.)

  • What filesystem mount-options are in us

/dev/sda3 on / type xfs (rw,noatime,nodiratime,seclabel,attr2,inode64,noquota)
/dev/sda5 on /home type xfs (rw,noatime,nodiratime,seclabel,attr2,inode64,noquota)
/dev/sda2 on /boot type xfs (rw,relatime,seclabel,attr2,inode64,noquota)
/dev/sda1 on /boot/efi type vfat (rw,relatime,fmask=0077,dmask=0077,codepage=437,iocharset=ascii,shortname=winnt,errors=remount-ro)

  • How are you copying files
  • I am using Redhat desktop nautilus

  • How are you moving files

  • I am using Redhat desktop nautilus

  • Is there a performance difference between copying and moving
    -No difference. Both causes the io wait more than 50%.

  • Are the source and destination on separate filesystems
    -No. I am coying and moving from / to /home directory. On my system they are on the same filesystem.

  • What's the underlying storage look like for the source/destination

  • It is a single disk. Hitachi HTS543232A7A384 (ES2OA90P)

  • Do you have A/V enabled on your system (e.g., I've seen McAfee absolutely murder performance on both copy and move operations)

  • No Anti virus enabled on my system

Oof... The HTS543232A7A384 is a slow hard drive to begin with (and likely accounts for a significant percentage of your performance issues).

Nautilus can (or at least used to) sometimes cause slower performance than work done at the CLI.

On your system, copying from "/" (hosted on sda3) to "/home" (hosted on sda5) is an operation that crosses filesystems. This means you won't get any kinds of performance improvments to moves that you'd get were you doing a move wholly within "/" or within "/home". That said, such 'same filesystem' performance improvements for moves are often casualties of graphical file-managers like Nautilus (which often implement a copy-then-delete rather than an actual move).

In general, XFS is optimized for large, sequential I/Os associated with streaming media. Its performance can get a bit dodgy when transfering lots of small files from filesystem to filesystem. XFS is particularly dodgy on delete operations ...which may be the real culprit here, as Nautilus may be waiting for the delete operations to complete.

At any rate, were I in your shoes, I'd probably try:
* Patching up the OS
* Try using the CLI to copy and/or move files and see if removing Nautilus from the equation improves things

Oooh... I see... Thanks for so much information... I have successfully patched the OS and updated all the packages installed... I tried to use CLI but the same 'hanging effect' is experienced. I read somewhere that disabling write-cache on the disk can improve the disk performance and also lessen the risk of data loss... This seems effective... My IO wait still reaches around 30% but there is no more hanging experienced...

Close

Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.