Hi all,
we are testing some settings on our ACS installation to get bandwidth up on
vm level. But it looks like we are stuck at 1Gbit.
Here is our test workflow:
* Create 2 instances within the same network
* Create a 20G file on one of the instances
* Scp the file from 1 instance to the other
* Overserve bandwidth via nload
We are using global setting vm.network.throttling.rate with value of 1000
and virsh dumpxml looks like this:
<interface type='bridge'>
<mac address=''/>
<source bridge='brbond1-673'/>
<bandwidth>
<inbound average='128000' peak='128000'/>
<outbound average='128000' peak='128000'/>
</bandwidth>
<target dev='vnet13'/>
<model type='virtio'/>
<link state='up'/>
<alias name='net0'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x03'
function='0x0'/>
</interface>
Which looks okay to us and we are able to get near max 1 Gbit bandwidth
during scp. (128000 kbyte/s = 1 Gbit)
When altering vm.network.throttling.rate to 2000 virsh dumpxml looks like
this:
<interface type='bridge'>
<mac address='02:01:00:d4:00:3e'/>
<source bridge='brbond1-811'/>
<bandwidth>
<inbound average='256000' peak='256000'/>
<outbound average='256000' peak='256000'/>
</bandwidth>
<target dev='vnet14'/>
<model type='virtio'/>
<link state='up'/>
<alias name='net0'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x03'
function='0x0'/>
</interface>
Which should be okay, it is double the size. But scp still maxes out near
1Gbit.
Any idea where to look for the bottleneck to get a 2 Gbit connection running
between 2 instances in the same network?
Regards,
Swen
