Dave Schneider,
Streaming video delivery is growing dramatically: according to the comScore Video Metrix, 5.5bn videos were viewed online in the UK during February 2010 (writes Dave Schneider, technical analyst at Ixia).

That's a rise of 37% over the same period in 2009, with YouTube.com continuing to account for the majority with 99.6% of all video views. Despite YouTube still dominating, the character of video viewing is changing as well, with more people watching longer content. This is reflected in cable provider Virgin Media’s recent launch of a broadband-based online movie rental service, where consumers can stream their choice of film on-demand to their computers.

There is also significant effort by broadcasters to make standard TV content available online. The BBC’s iPlayer replicates most BBC broadcast material and has been outstandingly successful: 120mn requests were serviced in January 2010, and according to Nielson time spent on the site increased by eighteen percent. Another success is US broadcaster NBC, whose coverage of the 2010 Winter Olympics included live and recently recorded content, complete with adverts.

As the usage increases, so too does the consumers’ expectation of the service. User satisfaction, often referred to as their quality of experience (QoE), is the ultimate measure of a reliable service. For steaming media services, users expect the following:

* Continuous play – without start/stop pauses
* Absence of video or audio skips
* Quick response to user actions that select and start the video, as well as pause, rewind and fast-forward operations
* Availability of low and high resolution versions


The challenge for service providers is achieving this QoE in a rapidly media landscape.

The streaming media distribution network

Video began as a limited service delivered over closed, private networks, but has moved quickly to be syndicated delivery over multiple networks. At the same time, video sources have broadened to include live and recorded material; free services have given way to paid services; and the ability to receive video content on multiple devices via multiple networks has become essential.

The other issue for service providers is that users want to view the content on a broad range of devices. Content is collected from disparate sources and distributed to devices and platforms including hybrid set-top boxes (STBs), personal computers, as well as Internet-connected smartphones and netbooks.

That said, the five stages of an aggregation/distribution network remain the same:

* Content sourcing – aggregation of content from physical media, live feeds and other sources.
* Content management – editing and management of the content, including uploading with digital rights management, with content encoded for multiple forms of delivery.
* Content hosting – centralised library of video content, plus replication to other levels of a CDN.
* Media delivery – the process of providing content in response to user requests, either through bulk download or via streaming.
* Player – the software application on the end-user device used to view and interact with content.


The most common network protocol used to transport video over IP networks is real-time streaming protocol (RTSP). RTSP is a stateful protocol used to establish and control media sessions between a media server and client viewer. RTSP clients issue VCR-like comments to control media playback. The transmission of the audio/video stream itself is most often handled by the real-time transport protocol (RTP), although some vendors have implemented their own transport protocol. RTSP and RTP are almost universally used to implement IPTV’s video on demand (VoD) features.

Streaming media players

The final part of the infrastructure is the application used to manage the content, but this adds another layer of complexity for service providers. Most video players, such as the Adobe Flash Player, use proprietary protocols that provide additional functionality and flexibility. Flash has an almost total presence on PCs and Macs, and is used to deliver over 80% of online videos.

But other applications such as Microsoft’s Silverlight are growing in popularity within the player market. Silverlight uses HTTP as its top-level transport mechanism and for media streaming. Using HTTP as a single transport mechanism can result in significant internal cost reduction for end-to-end delivery. A unique feature of Silverlight is adaptive streaming capability, which allows the player to adjust the stream playback quality based on real-time network conditions.

Where the reliability problems occur

All of these layers of complexity mean that jitter, loss, and latency are inherent in every IP network. These factors are most often compensated for by buffering at multiple network levels. Players commonly buffer data before beginning a presentation and read ahead to guarantee error-free delivery. Other techniques that are often employed in IPTV deployments are not suitable for Internet video delivery, including forward error correction (FEC) and periodic retransmission of lost segments.

There are many network levels, however, that must be transited between the streaming source and the destination – the core Internet, edge and aggregation networks, wireless networks, and enterprise LANs. These uncontrolled elements contribute to jitter and loss levels that cannot be compensated for by client-side buffering alone. So what can service providers do?

Stress testing for reliability

Rigorous testing of all streaming media delivery chain components is required to ensure user QoE, and components and networks must be tested under load to determine their limits.

It is especially important to test the devices that perform special handling on media flows:
Media servers – establish client connections, and convert and deliver content.
Content delivery networks – with sophisticated, multi-level architectures that distribute content from a central site to caching nodes and then finally to streaming servers located regionally and globally. Each level, and combination of levels, must be tested – especially for delay.

Data center components – the data centers maintained by service providers must balance their voice, video and data traffic to deliver QoE in all categories. Sophisticated devices, such as application delivery controllers (ADCs) that use deep packet inspection (DPI), inspect information flows to determine their required priorities and characteristics.

Wireless networks – 3G and LTE networks in particular are experiencing increased video traffic destined for mobile devices. Wireless network nodes must perform functions similar to those found in the data center – identifying flows and prioritising video and voice traffic over data traffic.

Pre-deployment testing is the only certain way of measuring maximum performance and true reliability at all load levels. This helps service providers to forecast the QoE they can achieve from their networks, taking into account all the potential reliability issues. Additionally, live network testing can be performed in low volume and QoE measurements can be made on individual streams to determine instantaneous network quality.

Conclusion

The largest part of Internet traffic growth will be associated with video delivery, and substantial infrastructure components will be purchased over the next decade to handle delivery of vast amounts of video content. Testing of the components and networks associated with that delivery are essential in order to bring quality services to market, properly scale the network and ensure end-user quality of experience.