[Gluster-users] improve performance
Roman Alekseev
rs.alekseev at gmail.com
Thu May 31 13:49:31 UTC 2012
GlusterFS shows the worst result:
Gluster NFS:
server96:/server18/ortoeuromag.ru on /glusterfs type nfs
(rw,mountproto=tcp,vers=3,addr=192.168.65.161)
time find . | wc -l
83949
real 75m5.243s
user 0m0.806s
sys 0m5.430s
Native client:
server96:/server18 on /mnt type fuse.glusterfs
(rw,allow_other,max_read=131072)
time find . | wc -l
83949
real 47m3.149s
user 0m0.904s
sys 0m4.320s
NFS without gluster:
time find . | wc -l
83931
real 0m13.420s
user 0m0.332s
sys 0m3.232s
vol info:
Volume Name: server18
Type: Distributed-Replicate
Status: Started
Number of Bricks: 15 x 3 = 45
Transport-type: tcp
Bricks:
Brick1: server96:/mnt/sdd5
Brick2: server29:/mnt/sdb5
Brick3: server89:/mnt/sdd5
Brick4: server96:/mnt/sdd6
Brick5: server29:/mnt/sdb6
Brick6: server89:/mnt/sdd6
Brick7: server66:/mnt/sdd7
Brick8: server29:/mnt/sdb7
Brick9: server89:/mnt/sdd7
Brick10: server66:/mnt/sdd8
Brick11: server29:/mnt/sdb8
Brick12: server89:/mnt/sdd8
Brick13: server96:/mnt/sda5
Brick14: server29:/mnt/sdd5
Brick15: server89:/mnt/sda5
Brick16: server96:/mnt/sda6
Brick17: server29:/mnt/sdd6
Brick18: server89:/mnt/sda6
Brick19: server29:/mnt/sdd7
Brick20: server89:/mnt/sda7
Brick21: server66:/mnt/sdb7
Brick22: server29:/mnt/sdd8
Brick23: server89:/mnt/sda8
Brick24: server66:/mnt/sdb8
Brick25: server29:/mnt/sdc8
Brick26: server89:/mnt/sdc8
Brick27: server66:/mnt/sdc8
Brick28: server96:/mnt/sdc5
Brick29: server29:/mnt/sdc5
Brick30: server89:/mnt/sdc5
Brick31: server96:/mnt/sdc6
Brick32: server29:/mnt/sdc6
Brick33: server89:/mnt/sdc6
Brick34: server66:/mnt/sdc7
Brick35: server89:/mnt/sdc7
Brick36: server29:/mnt/sdc7
Brick37: server29:/mnt/sda5
Brick38: server89:/mnt/sdb5
Brick39: server66:/mnt/sdd5
Brick40: server29:/mnt/sda6
Brick41: server89:/mnt/sdb6
Brick42: server66:/mnt/sdd6
Brick43: server29:/mnt/sda7
Brick44: server89:/mnt/sdb7
Brick45: server66:/mnt/sdb5
Options Reconfigured:
performance.io-thread-count: 32
features.quota-timeout: 600
performance.cache-size: 256MB
features.quota: off
auth.allow: *
I have 4 servers(server66, server89, server29, server96 ) and 2
dedicated servers(clients) which are connected to glusterfs storage.
Also could you please explain your advise with more details?
How to tune my storage?? Is someone who know how to configure this
distributed file systems correctly?
--
Kind regards,
R. Alekseev
More information about the Gluster-users
mailing list