The foundation of any successful Unified Communications deployment is a solid network. The network must have the capacity to support real time communications, while being cost effective, reliable, and secure. As UC evolves, how a network is designed must evolve too.
The drivers for changing the Network Architecture Paradigms are:
1) Embedded UC Apps--Real-Time voice and video are being embedded in more applications and devices. Separating this traffic is becoming more difficult with less value being returned for the effort. WebRTC will continue to drive this trend.
2) Video--Video is reaching critical mass and is becoming a requirement for organizations to support. Desktop, room, collaboration, training, and social media all use various forms of video. As HD video becomes mainstream, the demand for high capacity networks to support video will continue to grow.
3) The "Any" Requirement--The 24/7 flat-world economy reqiures networks and UC to be available anytime, anywhere, to anyone, on any device. Building special-purpose networks that only work for some people, in certain locations, on specific devices, is no longer preferred.
These drivers are changing some of the traditional Network Architecture paradigms for design, such as:
Separate voice/video transitioning to Classifying traffic as business critical or not--Current network architecture best practices include keeping voice/video traffic on its own VLAN and in its own QoS domain on the WAN. Going forward, a more effective model will be determining what traffic/applications are critical to an enterprise and prioritizing this traffic.
IT should classify their applications into two categories--those that keep the business running and those that help manage the business but can receive a lower priority. One solution is to divide the data center into two domains and prioritize traffic accordingly. Anything customer facing or core to the business operations, such as point of sale, customer service, manufacturing, or ERP, should be high priority. Email, software distribution, backups, file/print sharing, should be marked as low priority. If voice and video start to consume a large marjority--say, 60-80% of the network--then some type of classification within this group is required.
Tightly managing WAN transitioning to Overbuilding WAN--Cheap, best effort networks are more cost effective, with high enough reliability and performance to displace tightly managed networks. The explosion of consumer demand for high-speed Internet connections to support home and mobile entertainment and communication is bleeding over to the enterprise. The premise that WAN bandwidth is expensive and needs to be tightly managed is going away as more fiber and high speed last-mile technology is deployed. Ethernet access footprint now covers 85% of businesses in the U.S. and is 3-8x cheaper than traditional T1/DS-3/OC-X access. Cable and DSL offer additional coverage.
Separate Private & Public IP Networks Transitioning to One Network--Most carriers' core backbones carry both Internet and MPLS traffic, with the Internet traffic representing 5-10x more volume. If the carriers are moving to one network, enterprises should follow suit. Access has been the traditional bottleneck, but the carriers are overbuilding their access networks too. High speed aggregation networks are cheaper for the carriers to deploy and maintain than the traditional fixed-bandwidth SONET networks. As adaptive codecs are added to counter network jitter and dropped packets, the need for guaranteed bandwidth further erodes. Internet networks are 2-4x cheaper than MPLS networks with the same SLAs (assuming traffic runs end to end on the same carrier).
One final example: Which is better for an enterprise branch office--a 3Mbps T1/MPLS or a 20Mbps Ethernet/Internet VPN, if they both cost the same with the same SLAs? My recommendation is the latter, based in part on myown esperience, that my UC applications work better at home when I use my VPN than when I am in the branch office.
Network architecture requires developing solutions while dealing with competing requirements. The balance between cost, security, capacity, and reliability has always been a challenge. It will continue to be a challenge, but the assumptions that go into dealing with that challenge are changing.
