Hello there,

I am curious to see what are the plans with regards to integrate third
party SDN plugins, as it seems from OCP4 there's only one option, OpenShift
SDN.

I am in the process of taking a decision on a big SaaS player, so far these
are the pros and cons between VXLAN and BGP.

*# OpenShift default SDN Pluggin*
# Pros
1) Simple to Install.

# Cons
1) VXLAN will use CPU Cycles. It will Increase CPU usage on the host, and
high CPU usage will cause network performance to drop.

2) There should be no reason or need to run an overlay between 2 VMs in the
same Hypervisor, or between 2 Hypervisors in the same Rack, or between any
two points inside a Datacenter to be honest, unless really needed.

3) Difficult to troubleshoot

4) VXLAN is in essence will be a big Layer2 Domain, if we scale we will
exhaust the ARP table of those VMs. It can be increased, but how much?

*# Pure Layer3 (With Calico, or any other CNI Plugin).*
# Pros

1) Won't have any of the Cons listed for OpensShift SDN;
2) Real, routable IP address on the Containers;
3) Anyone with networking knowledge can troubleshoot a Networking
Connectivity issue that a Container may have, they will be seen as any
other VM or Baremetal from a Networking/Troubleshoot perspective;
4) BGP is well known by any Network Operator, even without OCP/K8s/Calico
knowledge a NetOps with BGP knowledge will be able to understand the setup;
5) BGP allow us to apply upstream policies if we want, based on
Communities.

# Cons
1) Slightly more complex to Install than default SDN Plugin.
2) RH may not support issues on calico itself if we have issues, but Calico
has support trough the Tigera Project (which also deliveries other cool
stuff for networking visibility at containers level).

Any advice, correction, suggestion or any clue about networking roadmap
would be highly appreciated.

Thanks for reading,

Regards,
_______________________________________________
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users

Reply via email to