Media files have been driving an increasing amount of traffic over mobile networks, and migration to faster networking technologies has dramatically increased the volume of encrypted video traffic. The traditional media delivery technology (Progressive Download) is failing to deliver acceptable quality of experience (QoE) at a high transmission rate. This has led to the introduction of the Adaptive Bit Rate (ABR) protocol. It can adapt the streaming bit rate to the available network bandwidth and restrict streaming quality to match the capability of the handset receiving the video. However, the ABR protocol does not work as well in mobile networks as it does over the internet. Mobile operators must, therefore, optimize ABR traffic.
A NetScaler appliance has unique capabilities to detect incoming video traffic and selectively optimize ABR videos.
How NetScaler Video Optimization Works
A NetScaler appliance can identify and optimize encrypted ABR traffic (including Facebook video traffic) over TCP, and YouTube ABR traffic over QUIC. The appliance has the following capabilities:
- Detect Progressive Download (PD) videos over HTTP.
- Detect and optimize ABR videos over HTTP.
- Detect and optimize ABR videos over HTTPS.
- Detect and optimize YouTube ABR videos over QUIC.
Also, the appliance uses the following support domains for detecting video traffic over TCP and QUIC protocols.
- Unencrypted ABR videos over TCP. Appliance detects all standard compliant video streaming web sites. The appliance detects ABR sessions by inspecting the response video payload header, URL, and HTTP headers.
- Encrypted ABR video over TCP. Appliance detects ABR sessions using a generic and heuristic algorithm based on domain, SSL header and traffic patterns. Using this, the appliance has a built-in support to detect top video web sites, with 95 percent accuracy and we continue to add support for new video types. NetScaler also has a program to provide additional verification for top encrypted ABR sites for a region or country to ensure network coverage.
- Encrypted ABR videos over QUIC. The appliance detects ABR sessions for QUIC based video provider, such as YouTube. The detection algorithm is on the basis of a heuristic leveraging the QUIC headers and domain. NetScaler will continue to add support for newer video sites using QUIC.
Optimizing the ABR video traffic can provide the following benefits:
- Manage the network during congestion in peak hours.
- Improve video play consistency and reduce video stalling.
- Enable new video service offerings (for example, Binge-on video services).
- Enable customers to select the best sustainable video quality.
- Provide a consistent user experience for the subscriber.
Video Optimization over TCP
NetScaler optimization of ABR traffic over TCP works as follows:
- HTTP or HTTPS traffic that the appliance receives over TCP is sent to the corresponding load balancing virtual server.
- The built-in detection policies bound to the virtual server combined with other proprietary detection algorithms evaluate the traffic.
- The policies use a set of built-in video detection signatures to detect the video type. The policy that matches traffic applies an action that categorizes the video type as one of the following:
- Clear-text PD
- Clear-text ABR
- Encrypted ABR
- The optimization policies bound to the same virtual server evaluate the traffic and determine the optimization bit rate to apply to the traffic.
- The optimization bit rate is applied if the traffic is either clear-text ABR or encrypted ABR.
A mobile service provider can improve the quality of experience (QoE) by setting the download speed for 2G, 3G, and 4G mobile traffic. This reduces the video start times or buffering events. Optimization can also reduce the amount of network bandwidth consumed by video sessions.
The optimization techniques include dynamic burst control and random sampling.
Dynamic Burst Control
NetScaler ABR optimization adapts dynamically to changing network conditions. It allows an initial burst rate of 1.3 times the configured pacing rate for 15 seconds. The initial burst rate applies to the beginning of every optimized ABR video session, even when multiple sessions use the same TCP connection or group of TCP connections.
The appliance also supports recovery bursts in the event that the bit rate supported by the network drops below the configured pacing rate. For example, if the effective bit rate drops at the 7th second and recovers at 15th second of the initial burst, the appliance recovers the loss during the next burst cycle. By doing so, the appliance dynamically optimizes the network bandwidth for all subscribers so that the quality of video remains consistent per pixel.
Note: When a recovery burst happens during an initial burst, the pacing bit rate must not exceed the maximum recovery-burst and initial-burst rates (you must not add the Recovery Burst factor on top of the Initial Burst factor). Otherwise, it might be so fast that the media player shifts to a higher quality mode. However, if necessary, you can extend the duration of the Initial Burst to compensate the unused bandwidth.
To estimate the savings from video optimization, the NetScaler appliance implements random sampling. With this technique, the appliance randomly selects a configurable percentage of the detected video traffic (the random sampling parameter is an integer number ranging from 0 to 100, so less than 1 percent is not possible). These randomly selected and non-optimized transactions (and sessions) become a reference group, and they are identified in the transaction logs (along with other characteristics, such as byte size and timer fields. The characteristics of the optimized sessions are also logged, and the reporting engine compares statistics of the optimized and reference groups to estimate the savings from optimization (including the savings from ABR Optimization).
Video Optimization over UDP
Google has introduced a new transport protocol called QUIC. Google’s QUIC protocol is very similar to TCP+TLS+HTTP/2 and is implemented on top of UDP. NetScaler can detect YouTube ABR videos streamed over the QUIC protocol, and apply ABR video optimization in a similar way as ABR over TCP.