MCUs & MPUs
Intelligent bandwidth optimisation
Intelligent bandwidth optimisation is vital if service providers are to combine quality of user experience with minimal operating expenses, as Richard Benson explains
M
obile service providers from around the globe are confronting massive traffic growth, as the popularity of smart devices and mobile broadband adoption drive dramatic increases in data traffic. In 2011 video traffic is thought to have exceeded all other mobile traffic so it must be managed without impacting other services that users have come to rely. Video requires relatively smooth delivery to look and sound good, making it not only a large contributor to network congestion, but less tolerant of congestion as well. Given the physical limitations of the wireless network, this translates into how best to reduce the bandwidth requirements of video while maintaining viable resources for the remaining services. It is clearly in every mobile service
provider’s best interest to find ways to alleviate the impact video content has on the network while still meeting consumer demand.
Mobile video consumption rates are outpacing revenues and to close that gap and convert growth into profitability, mobile service providers have to tackle two challenges: aligning revenue contributions with traffic consumption patterns and managing network congestion to improve customer satisfaction. Static provisioning has been used to manage and control services in legacy data environments. With this approach, provisioning systems like authentication, authorization and accounting are used to configure profiles. The profiles are applied just once, when the user establishes a data session. This static approach works for a one-size-fits-all, flat-rate model but video is a highly personalised service. To minimise operating expenses (OPEX), mobile service providers have to be able to make real-time adjustments at session start-up and during the middle of a session. Mobile service providers also have to react to different service levels, subscriber tiers, roaming and location status, network conditions, and applications.
28 April 2012
The increasing volume of traffic not only means congestion on the service provider’s network, it also affects their Internet Protocol transit costs. While the cost of bandwidth is declining, the growth in traffic more than makes up for it, driving overall costs higher for service providers.
Video coding standards overview Standards bodies, notably the Video Coding Experts Group (VCEG) and Moving Pictures Experts Group (MPEG), have developed a range of standards of varying computational expense with the key objective of reducing the transmission and storage requirements of video. Standards such as H.263, H.264, MPG2 and MPG4 have been widely accepted. As the adoption rate of video grows, then archiving video in legacy formats also grows, but these formats bring with them a couple of different challenges: 1. If an end-point does not support a codec, then there is a need to transcode it in the Content Delivery Network (CDN).
of available bandwidth. However, while the latest codecs may make more optimal use of bandwidth, there is significant scope for out-of-standard techniques to effectively retrofit advanced capabilities with older codecs.
Reducing video bandwidth requirements There are several approaches available to CDNs that can reduce the bandwidth resources needed to provide HDTV streams to the end-point. This will enable mobile service providers to trade the two key OAM drivers of compute expense (as a proxy for power consumption) against bandwidth consumption.
The fundamental driver behind the development of video and audio compression codecs has been to find efficient methods for storing and distributing video while maximising quality. The introduction of H.261, the latest variants of H.264 and emerging codec development have all helped make advances. Different codecs achieve different
compression ratios, but have an expense in compute resources needed for transcoding. Figure 1 shows, for a constant peak signal-to-noise ratio (PSNR), one measure of picture quality in the bit rate can vary by approximately two times while holding frame rate and picture size constant.
to achieve bandwidth gains. New techniques must be used to achieve bandwidth gains since there is limited room in the underlying codec.
Region of interest coding One particularly effective technique to reduce bandwidth demands is to intelligently code video based on where the consumer is likely to have attention focussed. This is known as Region of Interest (ROI) coding. For example, in a news story it is likely that the user is more focussed on the news moderator and less focussed on the background. This gives the service provider the opportunity to process the video stream to detect moving images like a person, then prioritise the coding of the person’s face and put less of a priority on the remaining picture. In practice this method is surprisingly effective. By its very nature, the processing seeks to preserve quality in areas the viewer is watching. Figure 2 and 3 shows how this may appear in real life. You can see there is no difference in the news anchor in the foreground, but the background has had high-frequency information attenuated (i.e., it appears a little more blurred). This can result in large savings of between 20 percent and 60 percent depending on content type and degree of coding. However, for other entertainment video other ROI techniques need to be employed to achieve bandwidth savings but will depend on other factors such as motion vector analysis. Motion vectors have been widely used in video codecs as predictors of where a particular pixel block will be in future frames, by analyzing past frames. The vectors produced can also be analysed to assist in prioritization of which parts of the picture warrant the highest quality encoding. At the simplest level, if a block is moving fast in a picture, then you should not be trying to encode it with high quality.
Figure 1
2. The benefits of the latest codecs are not available for the majority of video currently being consumed.
The first point has an added complexity in that while an end-point may well support an older video standard, the mobile service providers may well still wish to transcode the video to make better use
Components in Electronics
In the early days of mobile video, it was acceptable to scale picture size and frame rate to achieve successful delivery of the video stream. However, since the advent of the iPhone, smart phone manufacturers have chosen to compete on screen resolution and there is strong competitive pressure to not compromise on resolution
Managing compute expense efficiently Trading bandwidth for higher levels of optimisation comes at significant additional computational operating expense, and for the large number of streams necessary to make a significant impact this can rapidly scale to consume resources beyond the capabilities of standard x86 based CDN data centres. Therefore, multiple factors must be traded off:
www.cieonline.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64