Mobile Video Calling: Improving QoE
Today’s networks are “big” enough but not necessarily “fast” enough when it comes to mobile video calling. What are the key factors that influence quality of experience (QoE) for mobile video calls and are residential broadband networks ready for it?
Wanted: A better mobile video calling experience
Today, mobile video calling applications largely depend on video adaptation algorithms, advanced codecs on clients and an overlay peer-to-peer network to optimize QoE. But QoE for mobile video calling depends on more than the success of these techniques. It also depends on the network, competing traffic and the video “engine” in the client application. For example:
- Is there sufficient link bandwidth available for video calling?
- Is the network channel “stable,” and does it provide high throughput, low latency links?
- Does a prioritization scheme provide the necessary bandwidths among competing traffic flows?
- Are the video and audio codecs agile enough to operate with varying link bandwidths, with very low bandwidths?
- How quickly can client applications process video streams in real time?
A recent Bell Labs study of residential networks explored the factors that influence QoE in mobile video calls. The study found that network throughput capacity is ample for video calling. But huge queuing delays impair QoE for highly interactive applications. In other words, today’s networks are “big” (that is, high throughput capacity) enough but not necessarily “fast” (response time) enough for a truly great mobile video calling experience. The Bell Labs study found that some of the factors listed above have a major effect on QoE for mobile video calling, while others have almost none at all.
Significant issue: Link bandwidth
For video calling applications that generate large traffic volumes, maintaining high throughput on both downstream and upstream links is critically important. In today’s asymmetrically provisioned networks, the upstream link becomes the bottleneck in video calls. As shown in Table 1, major residential broadband networks in the U.S. provide greater speeds on downlinks than on uplinks. It is interesting to note that Long Term Evolution (LTE) mobile broadband networks now offer uplink speeds that exceed those of Digital Subscriber Line (DSL) residential broadband networks. With these speeds, the news of LTE smartphones with integrated Skype™ mobile video calling is not surprising.
Link speed stability also affects mobile video calling applications. Studies have found that high variations in residential broadband link speeds over the short term create challenges for mobile video calling applications, even if speeds remain stable over the longer term.
Significant issue: Video adaptation to changes in link bandwidth
In addition to the type of video and audio codec used, today’s video calling applications primarily distinguish QoE through video adaptation algorithms — bit-rate control and Forward Error Correction (FEC) mechanisms — that react to network congestion and random loss. Different reactions lead to different quality of experience. Most bit-rate control mechanisms reduce the transmission rate when congestion is detected and optimized for the available bandwidth. Generally, applications consider packet loss as a sign of congestion. However, in wireless networks packet loss can occur randomly due to signal fade, radio frequency (RF) interference and other channel impairments. If the application interprets all packet losses as congestion, it will incorrectly reduce its transmission rate during random loss, hence lowering QoE. A “smarter” mobile video calling application distinguishes between congestion and random loss. To make this distinction, the application must monitor packet delay and packet loss. During congestion, packet delay should spike, while with random loss packet delay should not change. To avoid packet loss regardless of link conditions, the transmission rate should always adjust promptly to changes in bandwidth. In addition, when a minimum bit rate is reached, a video call should be dropped (audio only) to prevent artifacts due to high packet losses. Theoretically, video codecs should adapt to bandwidth changes by supporting any requested bit rate. This “coarseness” of bit rates provides the flexibility to adapt to highly variable links. It is key to making the best use of available link bandwidth — especially with modern codecs like H.264 where there is a high correlation between frames.
Significant issue: Packet latency
For two-way, real-time mobile video calling, the packet latency, or delay, on upstream links is particularly critical. Packet delay is comprised of:
- Propagation delay: The amount of time it takes for a signal to travel from sender to receiver.
- Transmission delay: The amount of time required to push all the bits in a packet into the wire.
- Queuing delay: The wait time due to queues in the network to avoid packet dropping.
Queuing delays are typically the largest contributor to total packet delay. Studies have shown that queuing delays can reach 600 ms in DSL networks and 2 or more seconds in cable networks (due to large queues in some of these residential access networks). For interactive applications, the ITU G.114 standard recommends a maximum one-way delay of 150 ms because most users notice a round-trip delay of 250 ms or more. As a result, ideally, mobile video calls should not exceed 250 ms for end-to-end packet delay. Jitter can also affect packet delay and must be taken into account. Jitter can cause adaptive protocols that interpret Round Trip Time (RTT) changes as a sign of congestion to begin congestion avoidance too early. Recent Bell Labs research has suggested there are “buffer bloats” in many network elements, including operating systems and hosts, Network Interface Cards (NICs) and routers. Together, these “bloats” can cause significant delays in the network despite continued increases in bandwidth capacity. Any large upstream queue delays will impair mobile video call QoE.
Significant issue: Competing traffic flows
With more and more connected devices, and the popularity of video streaming applications such as Netflix and YouTube, the impact of competing traffic on mobile video calling QoE is expected to become a dominant challenge. A typical video calling application simply relies on network variables — available bandwidth, packet loss, and jitter — to throttle video and audio bit rates. It has no visibility of other applications that are competing for shared network resources. To a mobile video calling application, competing traffic is reflected as lower available bandwidth. With no centrally controlled prioritization mechanism, applications simply compete for available bandwidth. That means a video calling application with a good video adaptation algorithm can actually penalize itself. As it adapts its bit rate for the bandwidth, other applications consume any available bandwidth. Controlled prioritization mechanisms to “manage” competing traffic can be client-based, network-based, or a combination.
- In a client-based approach, the mobile operating system (OS) platform gives priority to a certain application and its traffic over others for a given set of use cases. For example, the Apple® iPhone® and Microsoft® Windows® Phone 7 support multi-tasking in a “controlled” fashion. Multiple applications can run at the same time, but the OS gives priority to applications, such as VoIP, over others. And not all applications can run simultaneously.
- In the network-based approach, network bandwidth allocation and traffic prioritization can be done at the home router or home service gateway — based on user specifications and real-time application requirements, for example.
- In a combination approach, competing traffic is prioritized by the client and the network. This approach provides prioritization control down to a specific device for a given home network. For example, traffic from a video calling application connected through a TV may be given a higher priority than traffic from that same application on a smartphone.
Secondary concern: Video and audio codec performance
Although there are numerous video codec standards, the most common ones used on consumer electronic devices are H.264, On2 VP7, and likely WebM on future Android devices. H.264 is the most popular codec for mobile video calling applications today. H.264 is reported to provide 50 percent bit-rate savings. That means it can deliver the same quality at half the bit rate compared to previous generation MPEG-2, H.263 or MPEG-4 Part 2 codecs. With the scalability and compressed bit rates that can be achieved with modern codecs, they are becoming a minor issue in QoE evaluation.
Minimal effect: Packet loss
Most DSL and cable networks have very low packet loss rates — typically below 1 percent, and 3G/LTE networks are engineered for low packet loss rate (typical target of 0.01 percent to 1 percent depending on the application). In addition, some DSL uplinks actively manage queues by proactively dropping packets using congestion avoidance algorithms such as Random Early Detection (RED). Packet loss is likely an insignificant factor in QoE for mobile video calls in residential broadband networks, considering these networks show remarkably low loss rates.
Minimal effect: Video processing power
Rapid innovation in mobile processor technologies means most smartphones and tablets now feature 1 GHz processors with built-in hardware graphics and video acceleration. They also feature the latest and fastest video and graphics processing chips. With the trend toward multicore processors with integrated Graphics Processing Units (GPUs), the video processing power on client devices is not an issue for mobile video calling. The only consideration is the impact of video processing on battery life.
Toward mass-market adoption
The network obstacles currently limiting QoE for mobile video calling can be addressed. Immersive applications, such as mobile video calling, will require new network designs and smarter algorithms to manage competing traffic on access and home networks. A home service gateway that can prioritize traffic by pre-allocating and dynamically adjusting link bandwidths based on user preferences could also improve QoE for mobile video calling. For the wireless network, the mobile video calling service will likely require mapping to a quality of service (QoS) Class Identifier (QCI). It will also require priority weighting or guaranteed bit rates (GBRs) in base stations. The Alcatel-Lucent Open API QoS (OAQ) solution provides this user-controlled prioritization by providing application programming interfaces (APIs) for the ACP client or server to specify which flow(s) get priority through the network. Moreover, the use of Data Packet Inspection (DPI) using the application-aware functionality in the Alcatel-Lucent 7750 Service Router (7750 SR) and its Multiservice Integrated Service Adapter (MS-ISA) blade can provide another helpful way to mark and prioritize traffic in the network. With these types of technology enhancements — along with adoption of new business models and resolution of regulatory challenges — mobile video calling has the potential to move into the mainstream. To contact the author or request additional information, please send e-mail to email@example.com.