[Linux-cluster] GFS: more simple performance numbers
Daniel McNeil
daniel at osdl.org
Tue Oct 19 23:44:06 UTC 2004
On Tue, 2004-10-19 at 11:05, Derek Anderson wrote:
> I've rerun the simple performance tests originally run by Daniel McNeil with
> the addition of the gulm lock manager on the 2.6.8.1 kernel and GFS 6.0 on
> the 2.4.21-20.EL kernel.
>
> Notes:
> ======
> Storage: RAID Array Tornado- Model: F4 V2.0
> HBA: QLA2310
> Switch: Brocade Silkworm 3200
> Nodes: Dual Intel Xeon 2.40Ghz
> 2GB memory
> 100Mbs Ethernet
> 2.6.8.1 Kernel/2.4.21-20.EL Kernel (with gfs 6)
> GuLM: 3-node cluster, 1 external dedicated lock manager
> DLM: 3-node cluster
> LVM: Not used
>
>
> tar xvf linux-2.6.8.1.tar:
> --------------------------
> real user sys
> ext3 tar 0m3.509s 0m0.262s 0m2.471s
> ext3 sync 0m1.051s 0m0.001s 0m0.204s
>
> gfs dlm 1 node tar 0m19.480s 0m0.474s 0m8.975s
> gfs dlm 1 node sync 0m3.167s 0m0.000s 0m0.195s
>
> gfs dlm 2 node tar 0m29.225s 0m0.465s 0m9.670s
> gfs dlm 2 node sync 0m3.398s 0m0.000s 0m0.224s
> gfs dlm 2 node tar 0m43.355s 0m0.562s 0m10.159s
> gfs dlm 2 node sync 0m4.922s 0m0.000s 0m0.226s
>
> gfs gulm 1 node tar 0m36.568s 0m0.491s 0m7.831s
> gfs gulm 1 node sync 0m3.243s 0m0.000s 0m0.152s
>
> gfs gulm 2 node tar 0m57.271s 0m0.527s 0m8.025s
> gfs gulm 2 node sync 0m2.471s 0m0.000s 0m0.145s
> gfs gulm 2 node tar 0m40.271s 0m0.482s 0m7.799s
> gfs gulm 2 node sync 0m3.636s 0m0.000s 0m0.224s
>
> gfs 6 gulm 1 node tar 0m32.671s 0m0.480s 0m7.480s
> gfs 6 gulm 1 node sync 0m3.436s 0m0.000s 0m0.120s
>
> gfs 6 gulm 2 node tar 0m38.130s 0m0.440s 0m6.920s
> gfs 6 gulm 2 node sync 0m3.731s 0m0.000s 0m0.110s
> gfs 6 gulm 2 node tar 0m58.564s 0m0.500s 0m6.670s
> gfs 6 gulm 2 node sync 0m0.971s 0m0.000s 0m0.070s
>
> du -s linux-2.6.8.1 (after untar):
> ----------------------------------
> real user sys
> ext3 0m0.103s 0m0.018s 0m0.055s
>
> gfs dlm 1 node 0m5.149s 0m0.041s 0m1.905s
>
> gfs dlm 2 node 0m11.127s 0m0.045s 0m1.839s
> gfs dlm 2 node 0m8.883s 0m0.033s 0m1.904s
>
> gfs gulm 1 node 0m0.355s 0m0.025s 0m0.330s
>
> gfs gulm 2 node 0m0.358s 0m0.024s 0m0.334s
> gfs gulm 2 node 0m0.358s 0m0.020s 0m0.338s
>
> gfs 6 gulm 1 node 0m0.314s 0m0.010s 0m0.290s
>
> gfs 6 gulm 2 node 0m0.308s 0m0.050s 0m0.250s
> gfs 6 gulm 2 node 0m0.303s 0m0.000s 0m0.310s
>
> Second du -s linux-2.6.8.1:
> ---------------------------
> real user sys
> ext3 0m0.074s 0m0.025s 0m0.049s
>
> gfs dlm 1 node 0m0.341s 0m0.027s 0m0.314s
>
> gfs dlm 2 node 0m0.325s 0m0.024s 0m0.300s
> gfs dlm 2 node 0m0.324s 0m0.020s 0m0.304s
>
> gfs gulm 1 node 0m0.354s 0m0.022s 0m0.332s
>
> gfs gulm 2 node 0m0.357s 0m0.023s 0m0.334s
> gfs gulm 2 node 0m0.359s 0m0.021s 0m0.338s
>
> gfs 6 gulm 1 node 0m0.299s 0m0.020s 0m0.280s
>
> gfs 6 gulm 2 node 0m0.299s 0m0.000s 0m0.300s
> gfs 6 gulm 2 node 0m0.302s 0m0.010s 0m0.290s
>
> rm -rf linux-2.6.8.1:
> ---------------------
> real user sys
> ext3 0m0.695s 0m0.013s 0m0.646s
>
> gfs dlm 1 node 0m10.056s 0m0.038s 0m4.720s
>
> gfs dlm 2 node 0m12.032s 0m0.043s 0m4.789s
> gfs dlm 2 node 0m13.803s 0m0.052s 0m4.787s
>
> gfs gulm 1 node 0m14.152s 0m0.066s 0m3.409s
>
> gfs gulm 2 node 0m12.408s 0m0.039s 0m2.355s
> gfs gulm 2 node 0m14.216s 0m0.038s 0m2.560s
>
> gfs 6 gulm 1 node 4m30.759s 0m0.140s 0m3.890s
>
> gfs 6 gulm 2 node 4m42.095s 0m0.060s 0m5.580s
> gfs 6 gulm 2 node 4m49.479s 0m0.140s 0m4.450s
>
What kind of raid setup is this (raid type, stripe size, etc)?
You are seeing some decent scaling, where I am not.
You have much faster processors.
Is the DLM code cpu intensive?
Daniel
More information about the Linux-cluster
mailing list