Open Stack

You can try either of the following methods to distributes the load of
processing received
packets and spreads them across multiple CPUs.

Receive Packet Steering:
Optimizing EC2 Network Throughput On Multi-Core Servers
<https://engineering.pinterest.com/blog/building-pinterest-cloud>

Multi-Queue virtio-net:
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Virtualization_Tuning_and_Optimization_Guide/sect-Virtualization_Tuning_Optimization_Guide-Networking-Techniques.html

On Tue, Sep 29, 2015 at 5:49 AM, Pedro Sousa <pgsousa at gmail.com> wrote:

> Hi all,
>
> I'm trying to deploy some nfv apps on my Openstack deployment, however I'm
> having some performance issues in my VMs, that start to lose UDP packets at
> a specific packet transmission rate.
>
>  Here's what I've tried and found so far:
>
> - VMs Centos 7.1 with 10GBe Neutron SR-IOV nics
> - Configured Memory Hugepages:
> http://redhatstackblog.redhat.com/tag/huge-pages/
> - Configured CPU Pinning and NUMA Topology:
> - Increased memory network buffers in kernel
> - Running "egrep ens6 /proc/interrupts" I see network Interrupts are not
> balanced evenly inside my guest across CPU cores, always hitting the same
> CPU.
>
> Concerning this last issue does anybody have some good advices on how to
> tackle this, how can I share the network load across the vcpus inside the
> guest or am I looking in the wrong direction?
>
> Some pointers would be appreciated.
>
> Regards,
> Pedro Sousa
>
>
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>


-- 
Best,

Jian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20151010/1b07160f/attachment.html>