OPNFV: How the Vendors are Approaching NFV


At the OPNFV Summit, there were many vendors and open source projects exhibiting their work. My prior blog discussed the talks at the conference. Although there is a common thread of using OpenStack and OpenDaylight, there are different approaches to providing higher level NFV functionality, ranging from PaaS, low level networking such as packet filtering and virtual switches, as well as hardware and everything in between. I list a few items that caught my eye.

  • Ericsson, which has been very active in OPNFV, showed a variety of its product capabilities as OPNFV.jpgwell as its involvement in the OPNV projects. One demo was their Continuum PaaS system, based on technology from Apcera, which they invested in. Their demo showed virtualized VNFs running within the Continuum PaaS as containers. The key takeaway for me is that the power of Apcera Continuum comes from its policy system, which affords high degrees of automation and provides for better security within the platform. Telco clouds have been associated with reliability, but have been under pressure to provide a more agile operational model, and a higher level PaaS such as Continuum enables a closer tie to a developer-centric model that works better for faster rates of innovation. How the developer-oriented world and the carrier grade infrastructure need to reconcile themselves is something to keep an eye on. Many other sessions were presented by Ericsson. Martin Bäckström (shown in photo) and others discussed how the open source model is fundamentally changing the way they develop systems.
  • HP was showing their carrier-grade Helion OpenStack. There seemed to be confusion stemming from Helion's withdrawal from the public cloud, but Helion OpenStack continues as the software foundation for HP’s NFV solutions. Many of the elements from their OpenStack come from their partnership with Wind River (an Intel company). Although many people associate OpenStack with the default reference implementation, since it’s a set of APIs, production ready editions are possible by slipping in alternative implementations. HP also has acquired companies such as ConteXtream, to fill out their NFV solutions, and the emphasis is on the software, as opposed to the HP hardware that can be part of the foundation.
  • Dell was also showing their NFV solution. Unlike HP, the emphasis was on hardware, and they have created starter-kits that are a starting point to perform PoC. They do have partnerships with firms like Red Hat, but in general, their approach is consistent with the rest of Dell, which is to provide choice on the higher level layers.
  • NTT DOCOMO and NEC showed the Doctor failure detection and notification system for NFV. For doctor.jpgtelcos, reliability is paramount, and operational capabilities have been slow in being introduced into an open source project such as OpenStack. To provide a better monitoring solution, NTT DOCOMO created the notification system which was demonstrated using an active/standby system running a workload, and we unplug a network cable (I actually did that in a live demo station), and the system automatically switches to the standby within a second, as opposed to minutes.  
  • Also shown were open source projects such as IO Visor, which appeals to telcos since it provides kernel level packet processing to improve the data plane. I wrote about IO Visor earlier, and I met Awais Nemat, PLUMgrid’s founder and chairman. He said that one item people often overlook is that IO Visor's architecture enables multiple IO modules to run side-by-side in the kernel, safely. This seems counterintuitive for those who understand how OS kernel level code works, but the key is how the IO Visor's virtual machines provide for safe isolation. The implication is that multi-purpose systems can run with high density, which translates into better CapEx. This is a new project in the Linux foundation, but given member support from Broadcom, Caviuum Cisco, Huawei, Intel, and Linux distros, SuSE and Ubuntu and Barefoot, a stealthy startup, it would be interesting to see how this technology will find its way into the platforms used throughout the network infrastructure.

Given that OPNFV is a new project, we're seeing a lot of innovation and attempts at new things. Some will stick but we will see varying schedules in which they reach maturity.

One last thing: There's an interesting take on whether virtual CPEs can be in containers or VMs. But that's for another blog.


partner economic validation


Topics: Networking