Three cardinal sins of the WAN in delivering UC are loss, latency and jitter
If you’re thinking UC, you’d better be thinking WAN optimization. WANs aren’t always architected to handle real-time apps (such as UC). The three cardinal sins of the WAN in delivering unified communications are loss, latency and jitter.
WAN optimization can control these — and in fact, is necessary: Most (80%) of companies running videoconferencing use some kind of WAN optimization. That said, WAN optimization isn’t a magic bullet. Even if you’re optimizing the WAN, you need enough bandwidth to start with. How much? VoIP traffic typically wants about 87Kbps per conversation, for example (you can drive it lower by choice of voice codec and trunking multiple simultaneous conversations). Video of reasonable quality wants about 768Kbps per video stream, and telepresence sessions want about 3Mbps.
Once you’ve got enough bandwidth, you need to prioritize, so that delivery of real time packets takes precedence over delivery of “best effort” traffic like e-mail.
Minimally, this means setting up class-of-service/quality-of-service, so voice and video packets get tagged for “real time” delivery, and routers and switches know to deliver real-time packets before anything else. This is where WAN optimization can help: Using WAN optimizers, IT can take finer-grain control of delivery of “nonpriority” traffic as well as of VoIP and video, and can match allocation of resources to applications’ (or users’) organizational importance. Optimizers that compress traffic and shape traffic rates keep space clear for VoIP and video, and gracefully manage rate reductions for best-effort apps when real-time traffic ramps up (something QoS doesn’t).
It is even possible to compress voice traffic! Lots of the bandwidth consumed by VoIP and video goes to packet headers; header compression and “packet train” aggregation techniques can reduce the overhead significantly.
And what about loss, latency, and jitter? All optimizers, by controlling other traffic and by gently shoving it aside during voice/video ramp-up, do a lot to make jitter a non-issue. Many optimizers mitigate packet loss problems through their own redundancy/recovery algorithms.
And, while no appliance can eliminate the real latency associated with speed-of-light limitations on long distance data transfers — you can’t cache a video conference, or spoof a voice conversation the way you can a handshake sequence between a file server and its client — some help with latency is possible.
Some optimizers can mitigate other sources of latency by artful manipulation of the rest of the traffic stream. After all, some latency comes from the process of getting packets onto the wire; an optimizer can make sure that all the other packets heading out are small enough to not make a voice/video packet wait too long at the router.
