[OpenIndiana-discuss] VERY slow server performance

[Roel_D]

Reading this it reminds me of the old days where IRQ's were important to
systems.
Those days my serial mouse could interfere with my modem. 

But I thought those days were way back.. 


-----Original Message-----
From: Ben Taylor [mailto:bentaylor.solx86 at gmail.com] 
Sent: donderdag 6 september 2012 16:56
To: Discussion list for OpenIndiana
Subject: Re: [OpenIndiana-discuss] VERY slow server performance

On Thu, Sep 6, 2012 at 12:48 AM, Stuart & Shirley <antarctic at sympatico.ca>
wrote:
>
> Hi - Look for some assistance in improving disk access performance on 
> my OpenIndiana install.
>
> Wanting to upgrade my server based on OpenSolaris and ZFS to 3TB 
> drives, I upgraded the OS to OpenIndiana 151a.
>
> Performance is terrible over the network as compared to the previous 
> OpenSolaris installation. Hardware is identical with the exception of 
> a second SATA controller (a second identical  Supermicro AOC-SAT2-MV8) 
> and additional hard drives.
>
> Perhaps related, the USB mouse is virtually none-operational. It will 
> only update the location of the mouse for about 0.5s out of every 10s. 
> This has not been an issue, as I usually connect remotely. But still 
> an indication that all is not well.
>
> Transfers on the  machine, from drive to drive appears about the same 
> as when running under OpenIndiana - but across the networks it's very
slow.
>
> Moving a large file from my windows 7 machine across the GbE lan - 
> transfers at ~11.5MB/s as reported by windows copy.
>
> Using zpool iostat, the write bandwidth is reported as high as 20M.
>
> Now if I run zpool iostat continuously (display every 3 seconds), copy 
> performance increases - moving into the 16 to 17 MB/s range as 
> reported by windows.
>
> Copying from one storage pool to another, zpool iostat will report 
> write bandwidths of 26M.
>
>
> My pool configuration is detailed below.
>
> Other's slow performance had been pegged to flow control enabled on 
> the Ethernet port - this was disabled. It did make a difference, but 
> not that dramatic.
>
> Any suggestions on how to improve performance and how to fix the mouse 
> would be greatly appreciated!
>
> Flow control properties:
>
> LINK         PROPERTY        PERM VALUE          DEFAULT        POSSIBLE
> e1000g0      flowctrl        rw   no             bi
no,tx,rx,bi
>
> Zpool configurations:
>   pool: rpool
>
> state: ONLINE
>
>   scan: resilvered 4.63G in 0h7m with 0 errors on Wed Jun  6 22:12:16 
> 2012
>
> config:
>
>      NAME          STATE     READ WRITE CKSUM
>      rpool         ONLINE       0     0     0
>        mirror-0    ONLINE       0     0     0
>          c3t0d0s0  ONLINE       0     0     0
>          c3t1d0s0  ONLINE       0     0     0
>
> errors: No known data errors
>
>   pool: tank_12T
>
> state: ONLINE
>   scan: none requested
> config:
>      NAME        STATE     READ WRITE CKSUM
>      tank_12T    ONLINE       0     0     0
>        raidz2-0  ONLINE       0     0     0
>          c3t7d0  ONLINE       0     0     0
>          c3t3d0  ONLINE       0     0     0
>          c3t5d0  ONLINE       0     0     0
>          c3t6d0  ONLINE       0     0     0
>          c5d0    ONLINE       0     0     0
>          c8d0    ONLINE       0     0     0
>
> errors: No known data errors
>
>   pool: tank_m
>
> state: ONLINE
>
>   scan: none requested
> config:
>      NAME        STATE     READ WRITE CKSUM
>      tank_m      ONLINE       0     0     0
>        mirror-0  ONLINE       0     0     0
>          c3t2d0  ONLINE       0     0     0
>          c7t6d0  ONLINE       0     0     0
>          c6d0    ONLINE       0     0     0
>        mirror-1  ONLINE       0     0     0
>          c3t4d0  ONLINE       0     0     0
>          c7t7d0  ONLINE       0     0     0
>          c4d0    ONLINE       0     0     0
>
> errors: No known data errors

The mouse interrupt issue is, in my experience, is not a new one.

I've been seeing this on solaris laptops for as long as I can remember.
The interesting issue for you is this only happened with the addition of
another sata controller.  This leads me to believe that there's some rather
confused logic in the interrupt handler, such that the mouse interrupts
basically gets preempted by the interrupts between the sata controller as
data is passed from one controller to the other.

Ben

_______________________________________________
OpenIndiana-discuss mailing list
OpenIndiana-discuss at openindiana.org
http://openindiana.org/mailman/listinfo/openindiana-discuss