XFS filesystem performance

5 pts.
Oracle 11g Database
Oracle Database
ZFS Filesystem
please advise some tips about how to create and tune a good performance of xfs filesystem to work effeciently with oracle database? thanks, nltt

Software/Hardware used:
SLES 11 SP1, HP DL785 with P8xx controllers

Answer Wiki

Thanks. We'll let you know when a new response is added.

First doing optimization, U have to analyze processes & processor Utilization, analyze memory utilization & performance and analyze network utilization and performance. Also consider the following, if u think it is necessary….
Change Network Environment viz., Monitor all other system components, Limit the collision domain, Check cable quality, Check both sides of a connection, Change network adapters, and Upgrade to a faster network type
One of the most important feature of XFS is that it is designed to handle a filesystem as large as a million terabytes. There r no of area, that helps to increase overall performance of SLES 11. Also I am strongly using SLED 11 as client and SLES 11 as server. And there r several options to optimize the filesystem as well as entire performance. XFS is enabled by default is disk quotas. Though u can assign quotas on a per-user, per-group, or per-directory basis for better performance.
Open/Edit /etc/fstab and add the <b>allocsize=X</b> option to the options list for the filesystem. U can insert the followings as ur criteria….
<b>atime, noatime</b>: These options set whether the access time of a file is updated in the inode (atime) or not (noatime). The option noatime should improve the performance.
<b>nodev, dev</b>: The nodev option prevents device files from being interpreted as such in the file system. To improve the FS choose nodev
<b>noexec, exec</b>: You can prohibit the execution of programs on a file system with the option noexec. It helps the same.
<b>nosuid, suid</b>: The nosuid option ensures that the suid and sgid bits in the file system are ignored. That depends how u use ur FS. Recommended nosuid.
<b>auto, noauto</b>: File systems set with the option noauto r not mounted automatically when the system is booted.
Now Open yast2 > system > /etc/sysconfig and point to system>File systems> Mdadm and change the value as MADAM_DELAY = 0 & MDADM_SCAN = no (If u r using RAID, then not to do) BOOT_MD_USE_MDADM_CONFIG = no & MDADM_DEVICE_TIMEOUT=0
Again, System>File systems> FAM> and change value as FAM_POLLING=no
<b>Kernel and syslog tuning….</b>
System>Kernel and change value as SCHED_MINTIMESLICE=1000, SCHED_MAXTIMESLICE=40000 & SKIP_RUNNING_KERNEL=yes
Again, System> UDEV> and change value as, UDEVD_MAX_CHILDS=64 & UDEVD_MAX_CHILDS_RUNNING=32
Again, Network>Hardware>config and change value as USE_SYSLOG=no
Add the hostnames in /etc/hosts
Open /boot/grub/menu.lst and add the word elevator=deadline where it starts kernel /vmlinuz…. and just before showopts (take extra care, when u do that)
If possible disable unnecessary services that r running in background. But be careful to disable them…. it entirely depends on ur system.
echo 256 > /sys/block/device/queue/read_ahead_kb
Now add the following parameter in <b>/etc/sysctl.conf</b>:
kernel.shmmax = 2147483648
kernel.shmmni = 4096 and kernel.shmall = 2097152
NOTE: Both r depend on each other.
<b>Setting Semaphores</b>:
kernel.sem = 250 32000 100 128 (i.e. semmsl, semmns, semopm, semmni respectively)
kernel.msgmni = 2018 (default is 16)
For file handling…. fs.file-max = 65536
<b>TCP/IP Tunning</b>:
net.ipv4.tcp_sack = 1
net.ipv4.tcp_window_scaling = 1
net.ipv4.tcp_rmem = 4096 256960 16777216 (space r required)
net.ipv4.tcp_wmem = same as above
net.ipv4.tcp_no_metrics_save = 1
net.ipv4.tcp_keepalive_time = 900 (default is 7200)
net.ipv4.tcp_ecn = 0
net.ipv4.ip_local_port_range = 1024 65000
net.ipv4.route.flush = 1
net.core.rmem_default = 262144 / 16777216 (take any one)
net.core.wmem_default = same as above
net.core.rmem_max = same as above
net.core.wmem_max = same as above
VM tuning is important part to increase performance of Oracle on SuSE Linux. For VM Tuning: http://mgogala.byethost5.com/Linux_VM.pdf
XFS has a tool that is called xfs_fsr tool and it can defragment a mounted partition without interrupting other work
The problem with XFS are
An XFS file system cannot be shrunk and Creation and deletion of directory entries can be much slower metadata operation than other file systems,depending on configuration
For more info about XFS, please visit here: http://oss.sgi.com/projects/xfs/


Discuss This Question:  

There was an error processing your information. Please try again later.
Thanks. We'll let you know when a new response is added.
Send me notifications when members answer or reply to this question.

Forgot Password

No problem! Submit your e-mail address below. We'll send you an e-mail containing your password.

Your password has been sent to:

To follow this tag...

There was an error processing your information. Please try again later.

Thanks! We'll email you when relevant content is added and updated.


Share this item with your network: