Re: [Vserver] Tuning HD performance

From: Herbert Poetzl <herbert_at_13thfloor.at>
Date: Mon 28 Aug 2006 - 23:22:01 BST
Message-ID: <20060828222201.GE27417@MAIL.13thfloor.at>

On Mon, Aug 28, 2006 at 02:40:38PM +0000, Arn?r Kristj?nsson wrote:
> Hi,
>
> I'm running a vserver running various LAMP servers. A while ago I
> installed munin to check how the server was behaving at various
> times. It turns out that the server has a LOT of iowait (up to 115)
> when under load.
>
> What I did was that I ran tiobench on the local HD and this is what I
> got:
> --
> Sequential Reads
> File Blk Num
> Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%)
> Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------
> --------- ----------- -------- -------- -----
> 2.6.14.3-vs2.0.1 100 4096 1 446.29 99.97%
> 0.006 0.04 0.00000 0.00000 446
> 2.6.14.3-vs2.0.1 100 4096 2 728.70 396.4%
> 0.007 0.05 0.00000 0.00000 184
> 2.6.14.3-vs2.0.1 100 4096 4 855.43 1587.%
> 0.011 0.20 0.00000 0.00000 54
> 2.6.14.3-vs2.0.1 100 4096 8 839.09 2328.%
> 0.013 56.08 0.00000 0.00000 36
>
> Random Reads
> File Blk Num
> Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%)
> Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------
> --------- ----------- -------- -------- -----
> 2.6.14.3-vs2.0.1 100 4096 1 418.79 96.48%
> 0.006 0.02 0.00000 0.00000 434
> 2.6.14.3-vs2.0.1 100 4096 2 754.47 386.3%
> 0.006 0.04 0.00000 0.00000 195
> 2.6.14.3-vs2.0.1 100 4096 4 839.15 1374.%
> 0.010 0.05 0.00000 0.00000 61
> 2.6.14.3-vs2.0.1 100 4096 8 739.47 870.8%
> 0.009 0.04 0.00000 0.00000 85
>
> Sequential Writes
> File Blk Num
> Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%)
> Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------
> --------- ----------- -------- -------- -----
> 2.6.14.3-vs2.0.1 100 4096 1 6.89 4.602%
> 0.021 9.83 0.00000 0.00000 150
> 2.6.14.3-vs2.0.1 100 4096 2 9.12 15.32%
> 0.031 80.06 0.00000 0.00000 60
> 2.6.14.3-vs2.0.1 100 4096 4 9.65 52.36%
> 0.040 0.54 0.00000 0.00000 18
> 2.6.14.3-vs2.0.1 100 4096 8 9.92 102.5%
> 0.066 100.08 0.00000 0.00000 10
>
> Random Writes
> File Blk Num
> Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%)
> Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------
> --------- ----------- -------- -------- -----
> 2.6.14.3-vs2.0.1 100 4096 1 0.34 0.148%
> 0.010 0.03 0.00000 0.00000 230
> 2.6.14.3-vs2.0.1 100 4096 2 0.34 0.352%
> 0.011 0.06 0.00000 0.00000 98
> 2.6.14.3-vs2.0.1 100 4096 4 0.37 1.033%
> 0.017 0.07 0.00000 0.00000 36
> 2.6.14.3-vs2.0.1 100 4096 8 0.38 1.596%
> 0.016 0.14 0.00000 0.00000 24
> --
>
> I'm wondering; shouldn't the CPU % be lower on reads?

well, I'm not sure what that data is trying to tell us,
because something like 2300 % CPU usage does not sound
like a sensible value to me, unless you have something
like 32 processors in that machine ...

> Is there a way to increase this performance?

sure, get better/faster disks, smarter controllers,
faster bus interfaces between cpu and i/o cards ...

sometimes even fine tuning of existing hardware can
make a big difference, like using the proper driver
with the 'right' options and/or avoiding partitions
which block disk access (when used simultaiously)

really depends on the setup

HTH,
Herbert

> A.
> _______________________________________________
> Vserver mailing list
> Vserver@list.linux-vserver.org
> http://list.linux-vserver.org/mailman/listinfo/vserver
_______________________________________________
Vserver mailing list
Vserver@list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver
Received on Mon Aug 28 23:32:49 2006

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Mon 28 Aug 2006 - 23:32:55 BST by hypermail 2.1.8