Serving the same filesystem concurrently over multiple protocols (Was: Re: [zfs-discuss] Lustre and ZFS)

Sander Klein roedie at roedie.nl
Sat Oct 19 14:42:59 EDT 2013


On 19.10.2013 18:57, Gordan Bobic wrote:
> On 10/19/2013 01:09 PM, Durval Menezes wrote:
>> Hi Fajar,
>> 
>> On Oct 19, 2013 8:48 AM, "Fajar A. Nugraha" <list at fajar.net
>> <mailto:list at fajar.net>> wrote:
>>  >
>>  > On Sat, Oct 19, 2013 at 6:02 PM, Durval Menezes
>> <durval.menezes at gmail.com <mailto:durval.menezes at gmail.com>> wrote:
>>  >>
>>  >> Hi Uncle,
>>  >>
>>  >> On Oct 19, 2013 7:26 AM, "Uncle Stoatwarbler" 
>> <stoatwblr at gmail.com
>> <mailto:stoatwblr at gmail.com>> wrote:
>>  >> >
>>  >> > (Yes, I know this should be a separate topic)
>>  >>
>>  >> Done :-)
>>  >> [...]
>>  >>
>>  >> > NFS server should be moved out of the kernel and back into
>> userspace. That's not something I have any influence over.
> 
> There is already a userspace nfsd. I use it for exporting fuse based
> file systems. The CPU overheads are higher, but they are usually not
> dominant compared to network bottlenecks.
> 

I'm currently building a test system to see how ganesha performs. I 
will write a post here to let you all know.

@Gordan: Which nfsd are you using? And, how much data do you movie with 
it? I'm looking for something that will push in excess of 10Gbit/s.

Regards,

Sander

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.



More information about the zfs-discuss mailing list