Response:Remembering The Management Plane « Network Heresy

Nicira is attempting to explain why they have chosen to use Open vSwitch Database protocol to configure network devices. By network devices, I mean virtual switches because I read this article as “we are making our own standars & API for configuring devices”.

My question is why?

There is a requirement to configure network device itself, in a number of different ways. But why  choose to build your own method instead of using OF-Config, the ONF Standard. By using the existing consensus and participating in this process, Nicira/VMware could integrate with any other vendor & have a better chance of partnering with existing hardware suppliers.

What’s the point of a new protocol and process using OVSdb ?

It’s a regular EMC/VMware tactic to create overlay their own APIs and call them amazing, and then control access to their ecosystem through accreditation and “approved vendor lists”. There is a thin veneer of credibility around validating SCSI RAID controllers because the technology is so dodgy, but for an external API that configures a network device ?

This isn’t open or good for networking. So why do this ?

via Remembering The Management Plane « Network Heresy.

  • Geraint Jones

    Probably because the board is staffed by folks from the valley who have cut their teeth with companies that make proprietary the norm.

  • cryptochrome

    To make it proprietary = money.

  • afewell

    My Take … are we not creating a layer of abstraction between software and infrastructure? And abstraction means each layer gets to evolve independently. VMware and any software should be able to make requests to the network and even define a particular service they would like to receive … but it is up to the infrastructure how it would like to fulfill the request. The upper layer should not know nor need to know what the lower-level protocol controlling the hardware is. If my app wants to request a change in the physical network and sends the request, maybe underneath I have a controller that telnets in and scripts the change … while clearly a more intelligent substrate can provide much more robust services, there is no reason why I should be precluded from getting the most out of whatever infrastructure I have if it meets my needs as a consumer. OVSDB is right in the idea they should not dictate the underlying protocol – and this should have no bearing on using OF as the network substrate, its the best there is and will enable more robust network services … but their should be a clean API to interface with software, and clean means the upper layer can only tell the lower layer what service it wants, not how the lower layer provides that service.

  • Kyle Bader

    Nicira doesn’t want to care about the underlying infrastructure, it simply acts like a dumb bus for their overlay. If applications need to reprogram the overlay network they should speak to the Nicira controller northbound API. I think standardization is more important at the controller API. The southbound API of the virtual switches is just a means to an end. I imagine they created their own API because OF-Config didn’t exist and the existing “standards” leave much to be desired (SNMP/Netconf).

  • Derick Winkworth

    money, Greg. money.

  • Nick Weber