Ok, one thing I remembered after posting my entry yesterday :

- In fact I was running *2* bonnie on the filer, the first on /vol/vol1 and the other on /vol/vol2, so the 2900 K/sec on per char / Sequential Output is just wrong. Quickly, I think I had to double that number. In fact, just forgot these

Here is the method I will use :

My test host (xenon) will be on a 10/100 switch (DLINK 3216) with the filer (using the built-in NIC). There is also 5 or 6 servers on the switch but I don't think it will interact. I will mount the filer /vol/vol1 with NFS (v2) and do a bonnie -2000 (not -1000 cause the filer is a 512 MB one). This first test will test the HD stuff and the network stuff of the filer, but I think the bottleneck is the network between my host and the filer. Yesterday I got 3/5 MB / sec traffic, it seems that bonnie generate a lot of traffic, maybe it is the data to be written in the file. I will look at the source later.

To test only the HD stuff, I think I will create a 10/20 GB file on one volume and mv it onto the other. I hope there will be the least possible networking stuff involved in this second test to stress the HD.

I also updated the report to add a SCSI 2 HD I have at home and a IDE HD.

And it seems that my lwm screenshot was a lot more popular than the bonnie report :

9:56 [huge:pts/2] xenon ~% grep screenshot.png /var/log/apache/access.log | wc -l
     26
9:56 [huge:pts/2] xenon ~% grep report /var/log/apache/access.log | wc -l
     3