[Gluster-users] bonding question

[Alex Crow]

Yes, but even with rr it's still one tcp connection. At layer2 it gets 
distributed over multiple physical links. TCP doesn't care or notice 
(except for retransmissions as I mentioned before).

This is one advantage of iSCSI/FCoE/FC/SCSI etc in that you can use 
"multipath" which is transparent, scales per-link close to linear and is 
part of the storage protocol (ie multiple abstract paths between 
initiators and targets) rather than the network stack.

You could serve up iSCSI from files on a mounted via FUSE from a 
glusterfs cluster, which would enable multipath, but I've only ever seen 
a demo of this on YouTube and I was not convinced that on its own it 
would be crash-consistent or resistant to gluster split-brain. Anyone 
else that's tried this is welcome to put me right on this.

Cheers

Alex


On 29/09/14 15:10, Demeter Tibor wrote:
> Hi,
>
> I would like to use glusterfs as ovirt-vmstore.
> I this case one vm, that is running on one compute node will use only one tcp connection?
>
> Thanks
>
>
>
>
> ----- Eredeti üzenet -----
>>> Ok, I mean this is a network based solution, but I think the 100MB/sec is
>>> possible with one nic too.
>>> I just wondering, maybe my bonding isn't working fine.
>> You should test with multiple clients/dd streams.
>>
>> http://serverfault.com/questions/569060/link-aggregation-lacp-802-3ad-max-throughput/
>>
>> rr
>>
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users