[perf-discuss] non-zero service time w/o any I/O ?

2006-12-06 Thread Leon Koll
On my s10 u3 host connected to the FC storage with MPxIO and SVM I see the non-zero [b]asvc_t[/b] when there is no I/O. Is there an explanation of this puzzle? TIA, [i]-- leon[/i] extended device statistics r/sw/s Mr/s Mw/s wait actv wsvc_t asvc_t %w %b device

[perf-discuss] Re: Turbocharging an NFS server

2006-01-15 Thread Leon Koll
Spencer wrote: [i]>Did you attempt to increase the number of load generating processes on the client?[/i] I increased the number of load generating processes on the client to [b]384[/b] and got [b][u]24626 IOps[/u][/b] [i]x-posted to http://napobo3.blogspot.com/2006/01/turbocharging-nfs-server

[perf-discuss] Re: Turbocharging an NFS server

2006-01-13 Thread Leon Koll
Both CPUs weren't entirely consumed, there are spare CPU cycles left. I increased the number of processes to 72 and the no. of IOps increased. My goal was not to get the max. SFS numbers but to find the optimal server configuration (kernel and NFS parameters) for specific setup. On the storage sid

[perf-discuss] Turbocharging an NFS server

2006-01-12 Thread Leon Koll
[i]x-posted: http://napobo3.blogspot.com/2006/01/turbocharging-nfs-server.html [/i] When out-of-the box S10/x86 showed mediocre NFS performance (measured by the SPEC SFS benchmark) the following measures improved the results more than twice : 1.increase the dnlc size (set ncsize = 0x10 in /