As voice becomes just another application that runs on an organization's shared IT infrastructure, it is critical that this shared IT infrastructure minimizes delay and jitter. What separates voice from other enterprise IT applications is its real-time, interactive nature that is very susceptible to milliseconds of delay and jitter. The ITU recommends that round trip delay be less than 150ms, wherever possible.The graphic below represents the UC3 (unified communications, collaboration, and content) architectural framework. The media services layer focuses on services that are specific to the real-time nature of unified communications such as conferencing, recording, and transcoding.
As organizations consolidate their telephony into data centers using SIP trunks, all media services should follow. But then comes the question, should some media services be handled locally--such as conferencing-in order to save network bandwidth and provide users with the best possible experience? This is similar to the question of putting in a caching server at an office building versus caching in the network and/or data center.
It is my recommendation that organizations centralize all media services into their data center. Yes, this strategy requires more network bandwidth. But, by keeping technology out of the field, the TCO to support UC services is lower, the ability to add and change services is quicker, and the user experience is consistent.
Within data centers for UC applications, it will be critical for companies to provide:
* Physical servers for media--Virtual servers typically add too much delay/jitter for real-time applications. Physical servers work best at the media layer, and all layers above can use virtual servers if an organization chooses.
* Network Peering--Minimize network delay to mobile, Internet, and third party networks by peering with them. For example, if a business has a lot of employees, customers, or partners using Skype, peering with Skype will help minimize delay.
* Minimize Transcoding--Allow the end-points to negotiate what codec to use versus the network. Adaptive codecs can adjust to the network.
* Get Competitive IP Networking Rates--Interconnect data centers at telecom hotels to get cost-effective bandwidth to the Internet, other companies, private WANs, and MPLS services.
* SBC--Session Border Controllers are real-time application firewalls that provide security, do protocol conversion, and provide demarcation points for supportability.
Yes, voice and video are real-time applications, but this does not preclude them from being centralized and using the shared IT infrastructure.