4 This document outlines the buffering policy used in the GStreamer
5 core that can be used by plugins and applications.
7 The purpose of buffering is to accumulate enough data in a pipeline so that
8 playback can occur smoothly and without interruptions. It is typically done
9 when reading from a (slow) and non-live network source but can also be used for
12 We want to be able to implement the following features:
14 - buffering up to a specifc amount of data, in memory, before starting playback
15 so that network fluctuations are minimized.
16 - download of the network file to a local disk with fast seeking in the
17 downloaded data. This is similar to the quicktime/youtube players.
18 - caching of semi-live streams to a local, on disk, ringbuffer with seeking in
19 the cached area. This is similar to tivo-like timeshifting.
20 - progress report about the buffering operations
21 - easy (backward compatible) application notification of buffering
22 - the possibility for the application to do more complex buffering
28 +---------+ +---------+ +-------+
29 | httpsrc | | buffer | | demux |
30 | src - sink src - sink ....
31 +---------+ +---------+ +-------+
33 In this case we are reading from a slow network source into a buffer element
36 The buffer element has a low and high watermark expressed in bytes. The
37 buffer uses the watermarks as follows:
39 - The buffer element will post BUFFERING messages until the high watermark
40 is hit. This instructs the application to keep the pipeline PAUSED, which
41 will eventually block the srcpad from pushing while data is prerolled in
43 - When the high watermark is hit, a BUFFERING message with 100% will be
44 posted, which instructs the application to continue playback.
45 - When during playback, the low watermark is hit, the queue will start posting
46 BUFFERING messages again, making the application PAUSE the pipeline again
47 until the high watermark is hit again. This is called the rebuffering
49 - during playback, the queue level will fluctuate between the high and the
50 low watermark as a way to compensate for network irregularities.
52 This buffering method is usable when the demuxer operates in push mode.
53 Seeking in the stream requires the seek to happen in the network source.
54 It is mostly desirable when the total duration of the file is not know, such
55 as in live streaming or when efficient seeking is not possible/required.
57 * Incremental download
59 +---------+ +---------+ +-------+
60 | httpsrc | | buffer | | demux |
61 | src - sink src - sink ....
62 +---------+ +----|----+ +-------+
66 In this case, we know the server is streaming a fixed length file to the
67 client. The application can choose to download the file on disk. The buffer
68 element will provide a push or pull based srcpad to the demuxer to navigate in
71 This mode is only suitable when the client can determine the length of the
74 In this case, buffering messages will be emited as usual when the requested
75 range is not within the downloaded area + buffersize. The buffering message
76 will also contain an indication that incremental download is being performed.
77 This flag can be used to let the application control the buffering in a more
78 intelligent way, using the BUFFERING query, for example.
80 The application can use the BUFFERING query to get the estimated download time
81 and match this time to the current/remaining playback time to control when
82 playback should start to have a non-interupted playback experience.
87 +---------+ +---------+ +-------+
88 | httpsrc | | buffer | | demux |
89 | src - sink src - sink ....
90 +---------+ +----|----+ +-------+
94 In this mode, a fixed size ringbuffer is kept to download the server content.
95 This allows for seeking in the buffered data. Depending on the size of the
96 buffer one can seek further back in time.
98 This mode is suitable for all live streams.
100 As with the incremental download mode, buffering messages are emited along
101 with an indication that timeshifting download is in progress.
106 In live pipelines we usually introduce some latency between the capture and
107 the playback elements. This latency can be introduced by a queue (such as a
108 jitterbuffer) or by other means (in the audiosink).
110 Buffering messages can be emited in those live pipelines as well and serve as
111 an indication to the user of the latency buffering. The application usually
112 does not react to these buffering messages with a state change.
118 A GST_MESSAGE_BUFFERING must be posted on the bus when playback temporarily
119 stops to buffer and when buffering finishes. When percentage field in the
120 BUFFERING message is 100, buffering is done. Values less than 100 mean that
121 buffering is in progress.
123 The BUFFERING message should be intercepted and acted upon by the application.
124 The message contains at least one field that is sufficient for basic
127 "buffer-percent", G_TYPE_INT, between 0 and 100
129 Several more clever ways of dealing with the buffering messages can be used when
130 in incremental or timeshifting download mode. For this purpose additional fields
131 are added to the buffering message:
133 "buffering-mode", GST_TYPE_BUFFERING_MODE,
134 enum { "stream", "download", "timeshift", "live" }
135 - gives the buffering mode in use. See above for an explanation of the
136 different modes of buffering. This field can be used to let the
137 application have more control over the buffering process.
139 "avg-in-rate", G_TYPE_INT
140 - gives the average input buffering speed in bytes/second. -1 is unknown.
141 This is the average number of bytes per second that is received on the
142 buffering element input (sink) pads. It is a measurement of the network
145 "avg-out-rate", G_TYPE_INT
146 - gives the average consumption speed in bytes/second. -1 is unknown.
147 This is the average number of bytes per second that is consumed by the
148 downstream element of the buffering element.
150 "buffering-left", G_TYPE_INT64
151 - gives the estimated time that bufferring will take in milliseconds.
153 This is measured based on the avg-in-rate and the filled level of the
154 queue. The application can use this hint to update the GUI about the
155 estimated remaining time that buffering will take.
160 While data is buffered, the pipeline should remain in the PAUSED state. It is
161 also possible that more data should be buffered while the pipeline is PLAYING,
162 in which case the pipeline should be PAUSED until the buffering finished.
164 BUFFERING messages can be posted while the pipeline is prerolling. The
165 application should not set the pipeline to PLAYING before a BUFFERING message
166 with 100 percent value is received, which might only happen after the pipeline
169 An exception is made for live pipelines. The application may not change
170 the state of a live pipeline when a buffering message is received. Usually these
171 buffering messages contain the "buffering-mode" = "live".
173 The buffering message can also instruct the application to switch to a periodical
174 BUFFERING query instead to more precisely control the buffering process. The
175 application can, for example, choose to not act on the BUFFERING message with
176 100 percent fill level to resume playback but instead use the estimated download
177 time to resume playback to get uninterrupted playback.
183 In addition to the BUFFERING messages posted by the buffering elements we want
184 to be able to query the same information from the application. We also want to
185 be able to present the user with information about the downloaded range in the
186 file so that the GUI can react on it.
188 In addition to all the fields present in the buffering message, the BUFFERING
189 query contains the following field, which indicate the available downloaded
190 range in a specific format and the estimated time to complete:
192 "busy", G_TYPE_BOOLEAN
193 - if buffering was busy. This flag allows the application to pause the
194 pipeline by using the query only.
196 "format", GST_TYPE_FORMAT
197 - the format of the "start" and "stop" values below
199 "start", G_TYPE_INT64, -1 unknown
200 - the start position of the available data
202 "stop", G_TYPE_INT64, -1 unknown
203 - the stop position of the available data
205 "estimated-total", G_TYPE_INT64
206 - gives the estimated download time in milliseconds. -1 unknown.
208 When the size of the downloaded file is known, this value will contain
209 the latest estimate of the remaining download time. This value is usualy
210 only filled for the "download" buffering mode. The application can use
211 this information to estimate the amount of remaining time to download the
214 For the "download" and "timeshift" buffering-modes, the start and stop positions
215 specify the ranges where efficient seeking in the downloaded media is possible.
216 Seeking outside of these ranges might be slow or not at all possible.
218 For the "stream" and "live" mode the start and stop values describe the oldest
219 and newest item (expressed in "format") in the buffer.
225 Some defaults for common elements:
227 A GstBaseSrc with random access replies to the BUFFERING query with:
229 "buffer-percent" = 100
230 "buffering-mode" = "stream"
234 "format" = GST_FORMAT_BYTES
236 "stop" = the total filesize
237 "estimated-total" = 0
239 A GstBaseSrc in push mode replies to the BUFFERING query with:
241 "buffer-percent" = 100
242 "buffering-mode" = "stream"
246 "format" = a valid GST_TYPE_FORMAT
247 "start" = current position
248 "stop" = current position
249 "estimated-total" = -1
255 Buffering strategies are specific implementations based on the buffering
256 message and query described above.
258 Most strategies have to balance buffering time versus maximal playback
263 NON-live pipelines are kept in the paused state while buffering messages with
264 a percent < 100% are received.
266 This buffering strategy relies on the buffer size and low/high watermarks of
267 the element. It can work with a fixed size buffer in memory or on disk.
269 The size of the buffer is usually expressed in a fixed amount of time units
270 and the estimated bitrate of the upstream source is used to convert this time
273 All GStreamer applications must implement this strategy. Failure to do so
274 will result in starvation at the sink.
276 * no-rebuffer strategy
278 This strategy tries to buffer as much data as possible so that playback can
279 continue without any further rebuffering.
281 This strategy is initially similar to simple buffering, the difference is in
282 deciding on the condition to continue playback. When a 100% buffering message
283 has been received, the application will not yet start the playback but it will
284 start a periodic buffering query, which will return the estimated amount of
285 buffering time left. When the estimated time left is less than the remaining
286 playback time, playback can continue.
288 This strategy requires a unlimited buffer size in memory or on disk, such as
289 provided by elements that implement the incremental download buffering mode.
291 Usually, the application can choose to start playback even before the
292 remaining buffer time elapsed in order to more quickly start the playback at
293 the expense of a possible rebuffering phase.
295 * Incremental rebuffering
297 The application implements the simple buffering strategy but with each
298 rebuffering phase, it increases the size of the buffer.
300 This strategy has quick, fixed time startup times but incrementally longer
301 rebuffering times if the network is slower than the media bitrate.