With over half a million daily active users of our audio-video Optimization Pack for virtualized Skype for Business, we’ve learned a lot about the impact of the network on user experience. Repeatedly, customers told me that before they were able to achieve satisfactory call quality with Skype for Business, virtualized or not, they needed to correct various network deficiencies. Otherwise, users would resort to dialing in to Skype meetings from their phone rather than take advantage of VoIP and video.

Frankly, our own implementation of Skype for Business here at Citrix revealed some sites and network links with unacceptably high sporadic packet loss (would you believe 75% on occasion?) — a rather appalling situation that real-time audio-video made very noticeable. And many Skype for Business customers have needed to review the capacity of their network connections to branch offices, especially with rising user expectations for High Definition video. (See also CTX222553 Factors Impacting Video Quality with Skype for Business.)

Microsoft publishes the bandwidth guidelines for Skype for Business on TechNet. Bandwidth consumption when using the HDX RealTime Optimization Pack is generally consistent with non-virtualized Skype for Business. The HDX RealTime Media Engine supports the same audio and video codecs commonly used by Skype for Business, and obeys the bandwidth restrictions configured on Skype for Business Server. If the network has already been properly provisioned for Skype for Business traffic, the Optimization Pack will work with that bandwidth, as it adds very little network overhead.

Audio and video traffic in optimized mode flows out-of-band from the ICA virtual channels. HD video resolution peer-to-peer calls will consume 1 Mbps or more. Conference calls may require 2 Mbps of bandwidth for HD video. The only additional network traffic generated by the Optimization Pack is from the low bandwidth ICA virtual channel control interactions between the RealTime Connector for Skype for Business on the VDA server and the RealTime Media Engine on the user device, and the compressed logging data sent from the Media Engine to the Connector. This additional traffic amounts to under 25 Kbps of upstream ICA bandwidth and about 5 Kbps of ICA downstream bandwidth.

The main video codec used by Skype for Business and RealTime Optimization Pack is H.264. H.264 is also the primary video codec for Microsoft Teams, which will gradually replace Skype for Business Online for Microsoft Office 365 customers. H.264 supports a wide range of video resolution and target bandwidth values. Bandwidth usage for video will always be constrained by Skype for Business Server bandwidth usage policies. In specific call scenarios, actual bandwidth usage may be even lower depending on current bandwidth availability and client endpoint capabilities. Optimization Pack also supports the legacy RT Video codec for interoperability with earlier versions of Microsoft unified communications software. Bandwidth usage with RT Video is similar to H.264, but video resolutions achievable with RT Video are limited to VGA or below.

Audio codec usage depends on the call scenario. Since the Microsoft Skype for Business Audio-Video Conferencing Server does not support SILK or RT Audio, these codecs are used only on point-to-point calls. Conference calls use G.722. SILK offers comparable audio quality to G.722 while consuming less bandwidth.

In addition to the codecs used by the native Skype for Business client, the HDX RealTime Media Engine offers a “super-wideband” codec, G.722.1C, offering superior audio quality when both parties on a point-to-point call are using XenApp or XenDesktop. This codec consumes 48 Kbps of network bandwidth. Optimization Pack 2.4 does not support the ultra-low bandwidth Siren codec, the predecessor to G.722.1. It does support G.722.1, for interoperability with third-party systems, although Skype for Business does not support G.722.1.

The Optimization Pack automatically chooses the best audio codec that is supported by all participants on the call and fits within the available bandwidth. Typically, this means:

  • A call between two Optimization Pack users will use the super-wideband G.722.1C codec at 48 Kbps with very good audio fidelity.
  • A conference call will use the wideband G.722 codec at 64 Kbps (159.6 Kbps with IP header, UDP, RTP, SRTP and Forward Error Correction).
  • A call between an Optimization Pack user and a native Skype for Business client user will use the wideband SILK codec at 36 Kbps (100 Kbps with IP header, UDP, RTP, SRTP and Forward Error Correction).
  • When an Optimization Pack user makes or receives a PSTN call, one of the narrowband codecs will be used (G.711 at 64 Kbps or narrowband RT Audio at 11.8 Kbps).

One of the best ways of improving branch office connectivity for real-time audio-video is with network virtualization. With NetScaler SD-WAN, bandwidth can be increased very economically by augmenting an existing MPLS link with low cost Business DSL. WAN virtualization (vWAN) maintains high availability for critical business applications — even higher than an MPLS link alone — while providing the bandwidth needed for High Definition video conferencing. On top of that, NetScaler SD-WAN can optionally duplicate audio packets over two links, using whichever one arrives first, to overcome packet loss and maintain quality.

On a well-engineered network, Skype for Business can deliver audio quality that’s noticeably superior to regular phone calls. Furthermore, techniques such as Forward Error Correction, Packet Loss Concealment and SD-WAN Packet Duplication can overcome many challenging network conditions to deliver a high quality video conferencing experience that users will enjoy.

This quarter, I am transitioning to the NetScaler SD-WAN team, where my new mission will be to introduce unique integration between SD-WAN and Citrix Workspace to provide the best user experience for both Cloud and on-premises customers. In fact, I’ve written this blog post on a flight back from Denmark, where I delivered my first presentation on SD-WAN and HDX to the newly formed Danish Citrix User Group Community (DKCUGC) at their inaugural meeting at Hotel Legoland. I’m excited to have been given the opportunity to apply my ten-year background in HDX high definition experience technologies to this rapidly growing market space, and I’m very pleased with the ongoing collaboration between Citrix and Microsoft to deliver the next generation of Intelligent Communications capabilities to our mutual customers.

Derek Thorslund
Director of Product Management, Software-Defined Networking

Follow Derek Thorslund on Twitter.