Open Networking Summit 2016 - changing worlds

open networking summitAt the Open Networking Summit, many were talking about open-source, but there were some contrarian views which made the conversations interesting. So let's discuss that.

I also saw some surprises, where companies that I thought were firmly in one path were making bold moves in another direction.

Pros and cons of open-source 

At one of the keynotes, Peter Levine, a general partner at Andreessen Horowitz, said that open-source is not necessary the source of innovation, and that traditionally, mature products have become open-source (MySQL) but the original innovation had closed-source origins (Oracle, DB2, etc.).


However, talks by leaders of open-source projects such as Linux, ONOS, OpenDaylight and OPNFV were stressing how they are collaborating and driving innovation. I think the truth lies someplace in the middle.  

Levine did say that it's fine for companies to identify some components useful for a larger community and share that in open-source.

For example, Google takes some elements of their container orchestration system and puts it in open-source (Kubernetes). Or, in the networking realm, IO Visor from PLUMgrid.  

Another view of open-source was at a session hosted by Martin Casado (who just joined Andreessen Horowitz himself). Casado believes that open-source is conflated with innovation, and panelist Koponen (at a startup) said it serves as a good way to understand the issues and learn about problems. So they felt open-source will help networking innovations in the future. How it ties into the business model (sell support and professional services?) is still a big question.

I do believe that open-source components are critical to modern infrastructure. Whether entire turn-key enterprise solutions can be built upon that is still to be determined. With open-source projects, there are so many different competing abstractions and layers released by different projects, it is difficult to ensure tight integration unless some party works to create a curated edition.

As to open-source based services, Levine said that using open-source to create a service like GitHub is also a fine model. GitHub is free, but people pay GitHub to use it in a commercial setting. Levine was skeptical of having small purely open-source-based software firms expect to generate sufficient revenue solely from support and services. This was an interesting observation from a conference that is tied to open-source.

Evolution to higher levels of the stack

There were some other surprises. An early OpenFlow supporter, NEC has now introduced a product that supports OCP-inspired networking equipment. This is a change from when NEC worked to create OpenFlow extensions to make a system that worked effectively and overcome limitations in the protocol. To provide complete compatibility, it typically required an environment based on NEC equipment at both ends — at the controller and the switch.   

Now it is leveraging the evolution of the OpenFlow protocol and emphasizing the compatibility of its ProgrammableFlow controller with OCP-inspired switches. It'd be good see if this can create an ecosystem of switches that works with its controller.

Omar Baldonado of Facebook also had a keynote, in which he described the rise of open hardware platforms. This open hardware view augments the focus on open software of this conference. Some of this was covered in my prior blog on Open Compute Summit.

We see support of open hardware even at the low level of network processor support. For example, Pica8 is a vendor that creates the PicOS network OS. It supports processors from many firms such as Broadcom or Cavium. This is accomplished by its vASICS as a hardware abstraction layer, supporting different chipsets for its software. So now the emphasis of Pica8 is as network software company and not a vendor of switches as it was perceived to be.

These are changes that support openness in different ways, and moving its emphasis to higher parts of the infrastructure stack and more towards software. We see that NEC is touting its capabilities in the controller, and Pica8 is touting is ability as a network OS vendor. 

I am going to investigate these issues further as I revisit OpenFlow, which is working to overcome the limitations seen in its early days, and I will also examine the role of controllers (vs. non controller environments) in networking. These are controversial issues.

The conference also had a stronger emphasis on telcos and service providers, who are actively creating proof of concepts based on OpenFlow technology. Whether this is an inherent property of the OpenFlow technology, reflects the needs of telcos, or both, needs to be examined. I plan to write about this in a future blog.

Microsoft making strides

Finally, in the world of network virtualization, I was pleased to see that Microsoft has steadily improved its network overlay virtualization technology within Windows Server. It helps tie running containers within Windows and now makes it an alternative to running containers atop Linux.  

Elements such as Hyper-V network virtualization have been around for a while, but Microsoft has made strides in making it a full network virtualizaton system with good use cases for stitching together micro-services. This is worth keeping an eye on for enterprises, as it may become a viable alternative to those who only considered plain Linux (and perhaps OpenStack) as a host for containers. Support for Docker Swarm and Apache Mesos (open source projects) by Microsoft is a a surprise to those who considered Microsoft's platforms to be very Windows centric and allergic to those that originated elsewhere. It's a refreshing breeze.

ONS takeaways

So the world of open networking continues to evolve, as view of openness change, and people revisit early assumptions and use cases. It's good that software projects and vendors are capable of changing with the times to react to market changes. If the projects or products become ossified, it will lose its relevance and fit with the market. I think it's worth taking a careful view of what open-source means, and acknowledge that while it has made tremendous contributions to the world of software, we need to realize it is not the answer to everything. Just as Linux has taken many years to get to where it is today, we probably will see that type of journey in other projects.

Video transcript:

Woman: The following is an ESG On Location video.

Dan: Hello, this is Dan Conde of ESG. I went to the Open Networking Summit in Santa Clara, California. The summit had, certainly, a telco slant to it, but there were plenty of items to see for the enterprise as well. There was an interesting discussion, led by Peter Levine, a venture capitalist from Andreessen Horowitz, who said that open source really isn't a way for innovation to occur, and it typically takes mature products and makes it freely available with the open source licenses. On the other hand, there was a viewpoint from representatives from the Linux community, Opendaylight, OPNFE, and ONOS, all open source projects, who said that they're working together to create a lot of innovation. I believe the truth resides somewhere in the middle and it remains to be seen as new projects emerge out of the open source community to what extent they can create true innovations.

One surprise for me was how Microsoft was showing how Windows and Azure platforms could become a good platform for running micro-services and how its Hyper-V networking technology and its SDN system can be used to stitch together these systems. The basic technology isn't completely new, it's been around Windows' server for a while, but having it mature to this level and making Windows' server potentially a good competitor to existing platforms like OpenStack for running containers is actually quite interesting. So this is something that is worth keeping an eye on.

We also saw a keynote by Omar Maldonado, a networking head of Facebook, and he showed how the internal technologies developed at Facebook is being open-sourced and also being given to the Open Compute Project as a way for Facebook itself to use the technologies as it's adopted and built upon by third party partners. And eventually, those technologies can trickle its way down to the enterprise. This was shown last week at the Open Compute Summit, and it's good to see it reinforced at the Open Networking Summit this week as well. So this is certainly a development that could affect the entire networking industry, and it's worth looking out for.

software defined storage insight

Topics: Networking