In a previous post I argued against the “One Internet”. I claimed that for video conferencing to work all access options to the network must be supported and the network itself has to be media-aware.
That being said, one might get the impression that I believe that only a dedicated video conferencing network, one that connects (at least) the entire enterprise and offers a “clean” environment for our precious means of communication, can provide a worthy quality of experience.
And it’s not that I’m against that type of solution. Whether it is ISDN, IP or MPLS based, the benefits are clear: a video network is permanent and always ready, the quality of service (QoS) is guaranteed, it is easy to handle and maintain, etc.
But if we are heading towards mass deployment of video conferencing infrastructure, dedicated networks are going to give IT managers and corporations a great big headache, as a dedicated network:
- is hard to scale
- is another network to administrate (on top of the existing IP networks)
- is expensive (double the network – at least double the expenses)
- is not the Internet
Convergence is the Key
The solution to the obstacles above lay in a converged network, one that incorporates all the IP traffic, including video conferencing. The benefits, in terms of cost, are clear, as notes Ben Gibson, Senior Director at Cisco:
“The clear advantages of converged networks are improved costs and IT resource productivity”.
But can a video conference network reside alongside other IP traffic, without risking quality and experience? Is the “converged network” a futuristic dream, or can you base your video conferencing network on your existing IP infrastructure? I decided to consult the local expert – Yossi Bronstein, AVP Corporate IT & IS at RADVISION, to learn how it is done in a global organization such as ours, with extensive IP-based communication deployment, including video conferencing (as the shoe maker doesn’t go bare-footed…)
The 1st thing Yossi did was show me the corporate network topology (see above). The 2nd thing Yossi did was emphasis that this is the corporate video network as well. RADVISION, says Yossi, has chosen to base its IT infrastructures on a converged network, due to the obvious reasons: costs and IT resources. Therefore, although RADVISION employees are using video conferencing much more than the average worker, all the IP traffic in RADVISION, including video conferencing, runs on the same network.
Dos and Don’ts of a Converged Network
This, of course, does not mean that the video is treated like any IP data over the network. On the contrary – the secret to properly converge video with “other” IP data, says Yossi, is priorities. Priority should be given to any “important”, “sensitive” data that goes over the network, but with video this is crucial.
Prioritizing should be embedded in the network architecture from the beginning. For that purpose, the network should be architected and configured by professionals. You have to know what the requirements are; you have to know what you’re dealing with; and you have to have the tools to monitor and maintain the network.
In RADVISION, for instance, as it can be seen from the network topology, every branch is connected through both the public Internet and dedicated MPLS lines. In general, the IP traffic uses the Internet, but at any given time, for any reason, it can be shifted to MPLS to guarantee proper transfer.
Same prioritization takes place in every aspect of the network. This can be achieved technically in many ways, such as:
- ToS can be used to guarantee low delay, high throughput and high reliability across routers via IP precedence.
- Differentiated services (diffserv) can be used to manage the traffic and provide different levels of QoS.
- Traffic shaping can be used to optimize or guarantee performance, latency and bandwidth in the network.
These priorities are decided according to the traffic characteristics: port allocation, source and/or destination, data type, etc. As mail server synchronization is prioritized for bandwidth and reliability, so can – for instance – a video call originating from the CEO’s office. And video should get more bandwidth than audio. And video between branches (in a distributed MCU architecture) should get enough bandwidth for proper quality.
And so, while every employee in RADVISION is capable of making a video call (via their personal IP Phone or SCOPIA Desktop) or a video conference (using their virtual meeting room), and although video traffic is substantial during work hours, Yossi says that on average only 3% of the time any external interference is required. Maybe this can explain how for a big organization such as RADVISION one system administrator deals with the entire global network.
And there you have it – you can have one network, accessible to all, via different means, protocols and infrastructures, and still enjoy a great experience for your visual communications, if you know how to set your priorities and not treat every packet as equal. And that is true, I thought to myself after my meeting with Yossi was over, not just for the RADVISION network or any corporate network, but for the Internet as a whole.