1.1 Conceptual Overview

The Media Streaming Server (MSS) system includes protocols to transmit data packets that originate from downloadable and streaming audio, video, and other multimedia data files.

Concepts that are specific to the Media Streaming Server system are:

Digital Rights Management: Digital Rights Management (DRM) provides content providers with the means to protect their proprietary music or other data from illegal uses, such as the creation of unauthorized copies. DRM technology protects digital content by encrypting it and attaching to it usage rules that determine the conditions under which a user can play back the content. Usage rules typically limit the number of computers or devices that have access to the content, or limit the number of times that content can be played.

Note  Digital Rights Management (DRM) is an optional component. For relevant supported Windows versions, see [MS-DRM].

Encoder: An encoder is a tool that is used to capture audio and video files and streams, to digitize them, and to provide them to media servers for distribution. For more information on creating a broadcast, see [WM9CSEB] section 5.

The following figure illustrates the live broadcast configuration.

Figure 1: Live broadcast configuration

Encoders typically capture the video and audio streams from capture cards and recording devices. To capture from an analog source, such as a video tape, the computer requires a capture card that recognizes the analog stream and converts it to digital media information. The encoder then converts the digital media information to encoded media that can be efficiently transported as streaming media.

Live broadcast: A live broadcast is often used when viewers want to see and hear an important event as it is occurring. For information about on-demand versus live broadcasts, see [MSFT-WMSDG], "Distributing content".

Media player client: A media player client is usually the destination point in the Media Streaming Server system, and is typically designed for rendering the media streams.

On-demand broadcast: An on-demand broadcast is a re-broadcast of a live event or of any media file that is not time critical. In this case, users can request the stream when they want to watch it and can control the playback to meet their requirements. For information about on-demand versus live broadcasts, see [MSFT-WMSDG], Windows Media Services 2008 Deployment Guide.

Playlist: A playlist is a file or collection of content files that is designed to play in a specific order or a query that results in a list of content files that are designed to play in a specific order.

The following table describes various playlists that are used by the MSS system.

Playlist formats	Details
Media playlist files	Designed for audio-only files and often referred to as the MP3 playlist. The playlist can provide URLs to HTTP servers or media servers.
Advanced Stream Redirector files	Advanced Stream Redirector files are based on the Extensible Markup Language (XML) syntax and are designed specifically to provide URLs to content from media servers to the client. For more information on Advanced Stream Redirector files, see [MSDN-ASX].
Windows Media Player playlist	A playlist query that is based on Synchronized Multimedia Integration Language (SMIL) that only works on local content and is not used by MSS. Windows Media Player Playlist syntax is based on SMIL 2.0. Clients load the playlists and process them locally. The playlist provides URLs or paths to the files. The client then streams the individual content by using the MSS system protocols. See the W3C website [W3C] for the SMIL 2.0 specification. For more information on the Windows Media Player playlist syntax, see [MSDN-MediaPlaylists].
Server-side playlists	Server-side playlists are a query method that is used to generate a list of content to be streamed to the client. The server processes the playlist locally and then streams the content to the client by using the MSS system protocols. The client receives a new Advanced Systems Format (ASF) file header each time when the server transitions from one entry to the next in the server-side playlist. Server-side playlists are beneficial as they allow the server (or encoder) operator to inject new playlist entries, such as advertisements, into a live program.

Origin server: An origin server is a media server that publishes on-demand or live content.

Distribution server: A distribution server improves the scalability of the Media Streaming Server system. A distribution server publishes content that it received from another media server. The distribution server has to be networked to the origin server and have permission to stream from the origin server.

A distribution server publishes content that it received from another streaming source, such as another media server. The origin server is the source of the content that is being streamed by the distribution server. Clients then connect to the distribution server as if it were the origin server. Distribution servers are located between the origin server and the client in the content stream and therefore can perform load balancing. Distribution servers provide an easy way to reduce the client load on a media server because the client content requests are distributed to several servers on the network. Publishing via distribution servers is illustrated in the following figure.

Figure 2: Publishing via distribution servers

Proxy server: A proxy server is a dedicated computer that proxies data between the media player client and the server. If the server is acting as a caching server, the proxy server requests a stream from the origin server and allows multiple clients to stream the content. Therefore the origin server is limited to one network request. If the content is broadcast content, the content cannot be cached. In this case, the proxy server can create a split stream for the content. The proxy server receiving the stream from the origin server splits the stream to distribute to multiple clients simultaneously without increasing the requests to the origin server. Proxy servers fall into three categories:

Forward proxy server: The forward proxy server can retrieve information from another server on behalf of a client. Typically, a client is explicitly configured to use a specific proxy server, and when the client requests content, the proxy server connects to an origin server to retrieve the content.

Reverse proxy server: A reverse proxy server is a proxy server that is configured to service all client requests. For unicast broadcasts, a reverse proxy server can reduce the load on the origin server by streaming multiple unicast streams while receiving only one stream from the origin server. For on-demand content, a reverse proxy server can reduce the load on the origin server by caching the content from the origin server and streaming it to clients from its cache.

To the client, the reverse proxy server appears to be the origin server. This structure enables the origin server to be isolated from the clients. A reverse proxy server can increase the security of the streaming media system because the client never connects to the origin server directly.

Transparent proxy server: A transparent proxy server is a server that transmits data between the server and the client without any modification of the data. It is a forwarding service that the client is unaware of.

Packet-pair bandwidth estimation: Packet-pair bandwidth estimation is a technique that is used to estimate the bandwidth of a streaming media connection over the Internet.

To estimate bandwidth, the server sends two or more consecutive packets of highly entropic data. The client estimates the bandwidth by measuring the difference between the times that it receives the packets. This method is usually reliable; however, if the client traverses a Network Address Translation (NAT) firewall or proxy server, the packet-pair bandwidth measurement might be inaccurate. Packet-pair bandwidth estimation is supported by the following protocols: the Real-Time Streaming Protocol (RTSP) Windows Media Extensions (RTSP-WME), the Windows Media Server (MMSP) Protocol, and the Windows Media Streaming HTTP Protocol (WMSP), as illustrated in the following figure.

Figure 3: Packet-pair bandwidth estimation

Fast start:  Allows the media player to buffer at speeds higher than the bit rate of the content requested. This enables users to start receiving content more quickly. After the initial buffer requirement is fulfilled, on-demand and broadcast content streams at the bit rate are defined by the content stream.

Fast start also allows a distribution server to request the data from the origin server at a faster bit rate. The bit rate that is specified in the fast start protocol headers ensures that the distribution server has enough data buffered to meet its requirements and the requirements of the media player client. To enable fast start, the protocols use the following two headers or tokens to request fast start:

• Accelerate headers: The tokens that the client uses to request a higher transmission rate and duration from the server.

• Burst headers:The tokens that the client uses to request a higher transmission rate and duration from the server. The client that sends the request is usually an intermediate device that is relaying the request for another client.

Fast start is supported only by the Windows Media HTTP Streaming Protocol (WMSP) and the Real-Time Streaming Protocol (RTSP) Windows Media Extensions (RTSP-WME).

Advanced fast start: Advanced fast start is designed to minimize startup latency in the media player client. Startup latency is the period of time starting when a viewer requests a stream by using the player and ending when the content begins playing. The primary reason for startup latency is the delay caused by buffering on the media player client.

Advanced fast start enables the media player client to begin playing a stream before its buffer is full. As soon as the media player client receives a minimum amount of data, it can begin playing a stream while its buffer continues to fill at an accelerated rate—a rate that is faster than the encoded bit rate of the content. When the buffer is full, acceleration stops, and the media player client begins receiving data at the encoded bit rate.

For advanced fast start to work effectively, adequate bandwidth has to be available above the encoded bit rate of a stream. For example, if 1,200 kilobits per second (Kbps) of bandwidth is available for an 800 Kbps stream, the media player client can use an acceleration rate of 1.5 times the encoded bit rate. If no additional bandwidth is available, the player fills its buffer before it begins playing a stream, and no benefit can be gained from either advanced fast start or fast start.

Advanced fast start is used only by clients that connect to a unicast stream and is supported only by the Windows Media HTTP Streaming Protocol (WMSP) and the Real-Time Streaming Protocol (RTSP) Windows Media Extensions (RTSP-WME).

Unicast streaming: Unicast streaming is a one-to-one connection between the media server and a media player client, which means that each client receives a distinct stream. Only those clients that request the stream receive it. The server can deliver content as a unicast stream from either an on-demand or a broadcast publishing point. Unicast streaming offers the benefits of interactivity between the player and server. However, the number of users that can receive unicast streams is limited by the bit rate of the content, the speed of the server network, and the available server resources. The number of users that are served is directly proportional to the amount of available server resources and instances. Unicast streaming is illustrated in the following figure.

Figure 4: Unicast streaming

Multicast streaming: Multicast streaming is a one-to-many relationship between a media server and the media player clients receiving the stream. With a multicast stream, the server streams to a multicast IP address on the network, and clients receive the stream by subscribing to the IP address. All clients receive the same stream and do not have control of content playback. Because there is only one stream from the server regardless of the number of clients receiving the stream, a multicast stream requires the same amount of bandwidth as a single unicast stream containing the same content. Therefore, multicast streaming improves the scalability of the server-side resources. More clients can be serviced with fewer server resources. Multicast streaming is illustrated in the following figure.

Figure 5: Multicast streaming

Core transport protocols: The Media Streaming Server system includes a number of individual protocols that rely on the User Datagram Protocol (UDP) to transport protocol packets. UDP is an unreliable protocol—it does not guarantee the successful receipt of the packets. The packets can be dropped, can arrive out of order, or can be replicated. Other Media Streaming Server system protocols, however, use Transmission Control Protocol (TCP), a reliable protocol that guarantees that packets arrive in the order specified. Section 2.3 includes a table that indicates UDP and TCP support for the Media Streaming Server system protocols.