Bernd Schemmer wrote:
> Hi,
> 
>  >>I would do just the opposite. Allow the service domain to cache the 
> files, especially since an update to a file might be
>  >>followed with another one,

By opposite I meant in the service domain only. Don't use directio and 
allow meta data logging.

> And use directio in the Guest Domain? Enabling caching in the Control 
> Domain and in the Guest Domains is not that useful I think.

I'd say neither, but haven't tried it.

In general, you use directio when the application has direct knowledge 
of what it wants to do with data blocks, such as a DB server would, and 
you don't want to OS to cache.

For general guest domain file systems (i.e. the OS), I would go with the 
default options. It would be interesting to hear what performance 
improvements you see under different application loads and type with the 
different options.

I don't now about how directio affects things when you *are* running a 
DB in the guest and whether the device(s) you use for the DB tables 
should be mounted differently in either domain. In other words, if you 
do use directio in the guest for the DB table locations, whether you 
must also in the service domain, or whether that gets passed automatically.

Steffen

> regards
> 
> Bernd
> 
> 
> Steffen Weiberle wrote:
>> Bernd Schemmer wrote:
>>> What are the recommended mount options for filesystems with the image 
>>> files for the (boot) disks of the guest LDOMS in the control LDOM?
>>>
>>> Currently we use forcedirectio,noatime,nologging - are these okay 
>>> and/or are there other recommended options to improve the performance?
>> I would do just the opposite. Allow the service domain to cache the 
>> files, especially since an update to a file might be followed with 
>> another one, and insure that if the service domain reboots, it reboots 
>> as fast as possible and is not forced to possibly do a lengthly fsck.
>>
>> Steffen
>>
>>> How about using VxVM volumes for these filesystems - anything to 
>>> consider here?
>>>
>>> I could not find anything about this issue in the documentation.
>>>
>>> regards
>>>
>>> Bernd
>>> -- 
>>> This message was posted from opensolaris.org
>>> _______________________________________________
>>> ldoms-discuss mailing list
>>> ldoms-discuss at opensolaris.org
>>> http://mail.opensolaris.org/mailman/listinfo/ldoms-discuss
> 
> 


Reply via email to