OPNFV: Focus On Testing And Collaboration

In my last column, I focused on the increasing influence of open source communities on the development of networking. To illustrate that increasing influence, I highlighted an SDN-focused open source community, the OpenDaylight (ODL) Project. While ODL is well known, some of the more NFV-focused open source communities, such as Open Platform for NFV (OPNFV), are less well known, particularly to enterprise IT organizations.

Similar to ODL, OPNFV is a member of the Linux Foundation. When OPNFV was announced, the Linux Foundation declared that OPNFV would establish a carrier-grade, integrated, open source reference platform that industry peers will build together to advance the evolution of NFV and ensure consistency, performance and interoperability among multiple open source components.

In a conversation I had with OPNFV Director Heather Kirksey, she emphasized that OPNFV’s charter is broader than just building a platform that realizes the European Telecommunications Standards Institute’s architectural framework. She emphasized that OPNFV is committed to influencing upstream communities such as ODL, OpenStack, and Open vSwitch, and enabling both functional and performance testing.

Last  June, OPNFV released its first  software, code named Arno. In line with OPNFV’s charter, Arno enables end users to deploy virtual network functions (VNFs) on the platform to test functionality and performance. Arno also reflects OPNFV’s commitment to testing by providing an automated toolchain that allows upstream projects to do automatic builds and verification as they develop independently.

While ETSI champions the interest that communications service providers (CSPs) have in virtualizing network functions, the Open Networking User Group (ONUG) focuses on the corresponding interest that enterprises have. In a white paper entitled Open Networking Challenges and Opportunities, ONUG discussed the cost and complexity of managing a large number of Layer 4-7 network appliances, such as application delivery controllers, WAN optimization controllers and firewalls.

When I asked Kirksey what OPNFV offers  enterprise organizations, she said that since its inception, OPNFV intended to focus on more than CSPs. Kirksey added that the group is working with ONUG, including taking ONUG's  test cases and incorporating them into OPNFV's functional tests.

One of the key architectural questions facing an organization like OPNFV that's developing a services platform is whether to take a tactical (aka piecemeal) approach  or an architectural (aka big bang) approach. Taking a piecemeal approach means focusing development on a small number of use cases for which there is clear business value. In contrast, if an organization takes a big-bang approach to developing an NFV platform, it decides on an architecture and possibly on some key enabling technologies and products that the architecture will utilize in order to support any and all NFV use cases.

Kirksey told me that at OPNFV, the people developing the platform want to create one that works for all use cases. However, she added that building a platform for all possible use cases is difficult without using initial versions of the platform to support individual use cases. According to Kirksey, each use case places different requirements on a platform and by using initial versions such as Arno to support specific use cases, the developers can better understand the requirements of future versions of the platform and also find bugs in the current version.

OPNFV.jpg

Kirksey said OPNFV’s second release will be in February 2016 and will support a broad range of SDN controllers. She also said that she wouldn’t be surprised to see OPNFV get involved in the discussion of virtual machines vs. containers.  According to Kirksey, because OPNFV does systems integration and testing, it can look at different ways to develop the platform (i.e., VMs vs. containers) and  test out the various theories about the pros and cons of VMs and containers.

There is a lot to like about open source development, particularly if it means products get to market faster. There is also a lot to like about the efforts of an organization such as OPNFV to integrate open source solutions from multiple sources. However, having worked for both CSPs and enterprise IT organizations, I struggle to think that many organizations will take a solution from an organization such as OPNFV and put that into production without any formal support from a reputable third party.

Kirksey said organizations are attracted to open source solutions for a variety of reasons, including reduced vendor lock-in. However, she added that most organizations will want a vendor to package the complete open source solution and provide ongoing support. Her statement confirms my belief that in order for sophisticated open source solutions to be widely accepted, there needs to be an organization that supports the solution in a fashion similar to how Red Hat supports Linux.