What is a BNG? What about the vBNG?
For those who may be new to the BNG, just a quick summarization of what is a BNG and what it does. In a nutshell, the BNG terminates subscriber PPPoE or IPoE sessions, typically originating from fixed line customers. Imagine an ISP with a fixed-line subscriber base wanting to access the Internet – a public IP address needs to be allocated, their traffic must be metered and rate-limited, and basic security must be provided. This is what the BNG is for, it is a network function sitting in the ISP core, making sure subscribers can browse the web.
Traditionally, the BNG is a router built on specialized hardware with the additional BNG functionality. This brings familiarity to the ISP since this is usually a vendor with which they are already working with and therefore they already know how they should configure and troubleshoot the BNGs. I’m using the plural “BNGs” since those are typically multiple boxes which must be configured separately, and great care has to be taken care of when designing redundancy.
The 5×9 Networks vBNG
We at 5×9 Networks have worked with these kinds of solutions and thought we could make it better. We designed a virtualized solution running of commercial-off-the-shelf x86 hardware for which we think addresses the most important drawbacks when working with traditional BNGs. It consists of three components:
- vDB – virtual Dashboard, the single point of contact for all management, monitoring and configuration tasks
- vBC – virtual BNG Controller, component for the control plane (subscriber management, Radius and routing connectivity)
- vBF – virtual BNG Forwarder; user plane component
The biggest advantage of our solution is the centralized management; the idea is to manage the BNG as a service (BaaS) rather than as a collection of independent routers. The vDB is the only point of contact when configuring any number of vBC and vBF instances and we have automated all sorts of features which make life easier:
- Configuration is based on pre-defined templates and instances are configured and put into production automatically once they finish booting (zero-touch configuration)
- IP addressing is done only on the vDB; once the IP pools are provided, the system itself takes care of subnetting, allocating pools to instances, advertising, summarizing and other routing tasks
- In case of failure or different traffic requirements, the system can automatically upscale or downscale new instances, automatically configure them (see the previous two points) and put them into production
- All these features (and more!) are meant to simplify the overall BNG experience, saving precious time in case of outages due to built-in automation
The other question we often hear is the scalability related to x86-based hardware, can it handle big subscriber numbers and traffic loads. With the not-so-latest and not-the-most-expensive Intel Xeon Gold CPU we can get 1 Tbps of traffic for 500-byte packet size on a total of 240.000 subscribers. Not bad for a server which you can buy for around 15.000€ directly in the store (and for much less if you’re a big ISP). Add additional servers and/or site for redundancy purposes and you can terminate more than 1 million subscribers in a geo-redundant fashion for the price of a single router running on specialized hardware. You have more traffic on the same number of subscribers? Just buy an additional server, 5×9 licenses only the number of subscribers.
One neglected advantage of a virtualized device is its flexibility when adding new features. When it comes to a BNG running on specialized hardware, it is difficult to add new features quickly, especially when dealing with big vendors. With a virtualized solution, it is quicker since it’s a software product – you can develop a new feature, immediately try it out in the lab and modify it if it’s not what you expected. Shorter development and deployment cycles mean you can get bug fixes and new features in a fraction of the time compared with a traditional BNG.
5×9 vBNG form factors
We can understand the complexity of adding a virtualization layer when deploying a vBNG – you might not have experience deploying virtualization in production, don’t know what to look out for or just don’t have time for a proper education. This is why we offer the vBNG as a black box, you can install an x86 server as a regular hardware appliance and we take care of the rest. We will install KVM, deploy virtual machines on top of it and you manage the instances as a single BNG. If we install all three 5×9 vBNG components on the same server, it can be an appliance, ready for standalone deployment.
If you are already running your own virtualization platform (OpenStack, VMware or even Kubernetes), our vBNG components support it and we can integrate onto it. Give us privileges to deploy new instances, and our Dashboard becomes an orchestrator, upscaling and downscaling forwarding instances with regards to the number of subscribers, throughput and/or the number of packets per second.
Summary
No product is perfect, but we strive to make the most of it. We combined our experience deploying and operating physical BNGs with our software skills to create the best version of the BNG we could think of. Knowing how a BNG is important to the ISP, we use continuous testing to minimize the number of bugs (like I said, no product is perfect!) and provide the stability which is required from a BNG. Hopefully we’ve been able to debunk some myths related to the vBNG, but we require feedback to know if we’re on the right track. Please let us know if you have any comments or suggestions on our contact form.
Author: Branimir Rajtar, 5×9 Networks
