Hacker News Reader

Comment by kazen44 | Hacker News Reader

i would disagree with you there, especially because there is very little on the sdn front which matches NSX-T in terms of SDN capabilities, this is something in which vmware has been ahead, the only other people with the same capabilities seem to be hyperscalers.

View on HN

Take a look at Proxmox SDN features: https://pve.proxmox.com/pve-docs/chapter-pvesdn.html (some of it is still in beta, I think).

I think it comes pretty close - close enough for probably most but the very largest of users, who, I think, should probably have tried to become hyperscalers themselves, instead of betting the farm and all the land around it on VMware (by Broadcom).

the thing it is mainly missing is multi-tenancy self service. (ipam integration seems very nice though).

NSX allows you to create seperate clusters which hosts VM's which run the routing and firewalling functionality.

NSX-T and what hyperscalers do is essentially orchestration of things that already exist anyway. The load balancing in NSX is mostly just some openresty and Lua which as been around for quite a while. Classic Q-in-Q and bridging also does practically all of the classic L2 & L3 networking that tends to be touted as 'new', while you could even do that fully orchestrated when Puppet was the hot new thing back in the day.

Some things (that were created before NSX) may have come from internet exchanges and hyperscalers, like openflow, P4, and FRR, but were really not missing parts that were required to do software defined networking. If anything, the only thing you really needed for SDN was Linux, and the only real distinction between SDN and non-SDN was hardwired ASICs in the network fabric (well, not hard-hardwired, but with limited programmability or 'secret' APIs).

Reply on news.ycombinator.com