Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 12 Feb 2009 23:38:24 -0500
From:      Charles Owens <cowens@greatbaysoftware.com>
To:        Scott Long <scottl@samsco.org>
Cc:        freebsd-scsi@freebsd.org
Subject:   Re: Disk performance on ESXi with FreeBSD 7.1
Message-ID:  <4994F940.7070301@greatbaysoftware.com>
In-Reply-To: <4991C6D5.4050400@samsco.org>
References:  <4991BD19.1000409@sebster.com> <4991C6D5.4050400@samsco.org>

next in thread | previous in thread | raw e-mail | index | archive | help
This is a multi-part message in MIME format.
--------------090402000900060509070705
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
X-WatchGuard-AntiVirus: part scanned. clean action=allow

Scott Long wrote:
> Sebastiaan van Erk wrote:
>> Hi,
>>
>> I'm running FreeBSD on ESXi but I'm having serious issues with disk
>> performance, and I'm wondering if it might have something to do with
>> the scsi driver (or the virtual hardware not returning proper values
>> for its capabilities or something)..
>>
>> I have both a FreeBSD-amd64 and Linux Ubuntu 8.10-amd64 virtual
>> machine (8GB disk, 512MB RAM, 2-CPU) and run dbench on both of them.
>> The linux machine is out of the box, not optimized for vmware,
>> VMI/paravirtualization is off, as is VMotion. The results for dbench
>> <n> are as follows:
>>
>> <n>             1               2               4
>> freebsd         12.0009         13.6348         12.9402         (MB/s)
>> linux           376.145         651.314         634.649         (MB/s)
>>
>> Thus there is approx a factor 30 difference for dbench 1, and I
>> cannot imagine linux being that much faster just due to some
>> performance tuning kernel parameters.
>>
>> I tried both the VMware LSI Logic controller and the BusLogic
>> controller. Here is the relevant dmesg output of both:
>>
>> LSI:
>> mpt0: <LSILogic 1030 Ultra4 Adapter> port 0x1080-0x10ff mem
>> 0xf4810000-0xf4810fff irq 17 at device 16.0 on pci0
>> mpt0: [ITHREAD]
>> mpt0: MPI Version=1.2.0.0
>> da0 at mpt0 bus 0 target 0 lun 0
>> da0: <VMware Virtual disk 1.0> Fixed Direct Access SCSI-2 device
>> da0: 3.300MB/s transfers
>> da0: 8192MB (16777216 512 byte sectors: 255H 63S/T 1044C)
>>
>> BusLogic:
>> bt0: <Buslogic Multi-Master SCSI Host Adapter> port 0x1060-0x107f mem
>> 0xf4810000-0xf481001f irq 17 at device 16.0 on pci0
>> bt0: BT-958 FW Rev. 5.07B Ultra Wide SCSI Host Adapter, SCSI ID 7,
>> 192 CCBs
>> bt0: [GIANT-LOCKED]
>> bt0: [ITHREAD]
>> da0 at bt0 bus 0 target 0 lun 0
>> da0: <VMware Virtual disk 1.0> Fixed Direct Access SCSI-2 device
>> da0: 40.000MB/s transfers (20.000MHz DT, offset 15, 16bit)
>> da0: 8192MB (16777216 512 byte sectors: 255H 63S/T 1044C)
>>
>> Something that I noticed was the extremely slow transfer rates
>> mentioned with the da0 device.
>>
>> When I'm running dbench the server is not very busy:
>>
>> CPU:  0.2% user,  0.0% nice,  6.4% system,  0.7% interrupt, 92.7% idle
>>  1172 root        1  -8    0  4604K  1228K biowr  1   0:41  4.98% dbench
>>
>> I really want to get this working because I want to run a big
>> production site on FreeBSD. But currently the disk speed is just
>> unworkable.
>>
>> I was wondering if anybody had any ideas about how to get proper disk
>> speeds on FreeBSD, making it a viable guest operating system.
>>
>> If any other info is needed, I'm willing to invest quite some time to
>> provide it!
>>
>> Regards,
>> Sebastiaan
>
> Run the following command:
>
> sudo camcontrol tags da0
>
> If it returns something like this:
>
> (pass0:mpt0:0:0:0): device openings: 1
>
> then run the following command:
>
> sudo camcontrol tags da0 -N 64
>
> If this works in improving performance, it can be put into a startup
> script.  I have no idea why the controller is misbehaving with this yet,
> but I'm working on it.
>
> Scott
>
I'm also seeing poor performance (much less than linux), running within
ESX Server... I'm using FreeBSD i386, single CPU, 1 GB RAM.

I've run the "camcontrol tags da0 -N 64" command and have seen
significant improvement (dbench numbers boosted by factor of 5 or so).  
Not sure it's quite enough, compared to Linux... but a big help already
(thanks!).

I'd be glad to do more testing / poking if it could help get this
figured out.

Thanks,
Charles


**Charles Owens**
*Great Bay Software*

--------------090402000900060509070705--




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4994F940.7070301>