Amazon Kinesis Video Streams service quotas

Kinesis Video Streams has the following service quotas:

Important

The following service quotas are either soft [s], which can be upgraded by submitting a support ticket, or hard [h], which can't be increased. You will see [s] and [h] next to individual service quota in the tables below.

Control plane API service quotas

The following section describes service quotas for control plane APIs. TPS stands for transactions per second.

When an account-level or resource-level request limit is reached, a ClientLimitExceededException is thrown.

API	Account limit: Request	Account limit: Streams	Stream-level limit	Relevant exceptions and notes
CreateStream	50 TPS [s]	10,000 streams per account [s] in all supported Regions. Note This limit can be increased up to 100,000 (or more) streams per account [s]. Sign in to the AWS Management Console at https://console.aws.amazon.com/ and request an increase of this limit.		Devices, CLIs, SDK-driven access, and the console can all invoke this API. Only one API call succeeds if the stream doesn’t already exist.
DeleteEdgeConfiguration	10 TPS [h]	N/A	1 TPS [h]
DeleteStream	50 TPS [h]	N/A	5 TPS [h]
DescribeEdgeConfiguration	50 TPS [h]	N/A	5 TPS [h]
DescribeImageGenerationConfiguration	50 TPS [h]	N/A	5 TPS [h]
DescribeMappedResourceConfiguration	50 TPS [h]	N/A	5 TPS [h]
DescribeNotificationConfiguration	50 TPS [h]	N/A	5 TPS [h]
DescribeStream	300 TPS [h]	N/A	5 TPS [h]
GetDataEndpoint	300 TPS [h]	N/A	5 TPS [h]	Called every 45 minutes to refresh the streaming token for most PutMedia/GetMedia use cases. Caching data endpoints is safe if the application reloads them on failure.
ListEdgeAgentConfigurations	50 TPS [h]	N/A	N/A
ListStreams	50 TPS [h]	N/A
ListTagsForStream	50 TPS [h]	N/A	5 TPS [h]
StartEdgeConfigurationUpdate	10 TPS [h]	N/A	1 TPS [h]
TagStream	50 TPS [h]	N/A	5 TPS [h]
UntagStream	50 TPS [h]	N/A	5 TPS [h]
UpdateDataRetention	50 TPS [h]	N/A	5 TPS [h]
UpdateImageGenerationConfiguration	50 TPS [h]	N/A	5 TPS [h]
UpdateNotificationConfiguration	50 TPS [h]	N/A	5 TPS [h]
UpdateStream	50 TPS [h]	N/A	5 TPS [h]

Media and archived-media API service quotas

The following section describes service quotas for media and archived media APIs.

When an account-level or resource-level request limit is reached, a ClientLimitExceededException is thrown.

When a connection-level limit is reached, a ConnectionLimitExceededException is thrown.

The following errors or acks are thrown when a fragment-level limit is reached:

• A MIN_FRAGMENT_DURATION_REACHED ack is returned for a fragment below the minimum duration.

• A MAX_FRAGMENT_DURATION_REACHED ack is returned for a fragment above the maximum duration.

• A MAX_FRAGMENT_SIZE ack is returned for a fragment above the maximum data size.

• A FragmentLimitExceeded exception is thrown if a fragment limit is reached in a GetMediaForFragmentList operation.

Data plane API service quotas

API	Stream-level limit	Connection-level limit	Bandwidth limit	Fragment-level limit	Relevant exceptions and notes
PutMedia	5 TPS [h]	1 [s]	12.5 MB/second, or 100 Mbps [s] per stream	Minimum fragment duration: 1 second [h] Maximum fragment duration: 20 seconds [h] Maximum fragment size: 50 MB [h] Maximum number of tracks: 3 [s] Maximum fragments sent per second: 5 [h] Maximum fragment metadata limit: 10 tags [h]	PutMedia requests are streaming, long-running connections. You don't need to open a new connection for each piece of data because you can send multiple fragments in a single persistent connection. If you attempt more than one concurrent PutMedia connection, Kinesis Video Streams throttles the latest connections with a ConnectionLimitExceededException error message.
GetClip	N/A	N/A	100 MB size limit [h]	Maximum number of fragments: 200 [h]
GetDASHStreamingSessionURL	25 TPS [h]	N/A	N/A	N/A
GetHLSStreamingSessionURL	25 TPS [h]	N/A	N/A	N/A
GetImages	N/A	N/A	100 MB [h]	N/A	Maximum number of images per request is 100 [h]. Note The minimum value for SamplingInterval is 200 milliseconds (ms), which is 5 images per second.
GetMedia	5 TPS [h]	3 [s]	25 MB/s or 200 Mbps [s]	Maximum of 5 fragments sent per second [h]	GetMedia requests are streaming, long-running connections. You don't need to open a new connection for each piece of data because you can send multiple fragments in a single persistent connection. If you attempt more than three concurrent GetMedia connections, Kinesis Video Streams throttles the latest connections with a ConnectionLimitExceededException error message. If a typical fragment is approximately 5 MB, this limit means ~75 MBps per Kinesis video stream. Such a stream would have an outgoing bitrate of 2x the streams' maximum incoming bitrate. Note GetMedia isn't used for HLS/DASH playback.
GetMediaForFragmentList	N/A	5 [s]	25 MB/s or 200 Mbps [s]	Maximum number of fragments: 1000 [h]	Five fragment-based consuming applications can concurrently invoke GetMediaForFragmentList. Further connections are rejected.

Video playback protocol API service quotas

API	Session-level limit	Fragment-level limit
GetDASHManifestPlaylist	5 TPS [h]	Maximum number of fragments per playlist: 5,000 [h]
GetHLSMasterPlaylist	5 TPS [h]	N/A
GetHLSMediaPlaylist	5 TPS [h]	Maximum number of fragments per playlist: 5,000 [h]
GetMP4InitFragment	5 TPS [h]	N/A
GetMP4MediaFragment	20 TPS [h]	N/A
GetTSFragment	20 TPS [h]	N/A

Fragment-metadata and fragment-media quotas

Kinesis Video Streams APIs for accessing archived media are throttled based on the number of fragments requested rather than the number of API calls. APIs are rate-limited by both the number of fragment metadata and the number of fragment media that's requested. The fragment metadata and fragment media quotas are applied per stream. In other words, requests for fragment metadata or media in one stream don't apply to the quotas of another stream. However, within a given stream, each quota is shared across multiple APIs. This means that, for a given stream, requests for fragments across different APIs consume from the same quota. When either the fragment metadata or fragment media quota for a stream is exceeded, the API returns a ClientLimitExceededException. The following tables show how the APIs consume from each of the two types of quota. For the second column in these tables, assume that if a stream has a quota of N, that means the APIs have N points to consume from that quota type for that stream. The GetClip API appears in both tables.

Fragment-metadata quota consumption

API	Number of quota points consumed per request	Shared quota (N)
ListFragments	Value of the MaxResults parameter	10,000 quota points per second, per stream [h]
GetClip	Number of fragments in the resulting clip
GetHLSMediaPlaylist	Value of the MaxMediaPlaylistFragmentResults parameter
GetDASHManifest	Value of the MaxManifestFragmentResults parameter
GetImages	Value of 400 + max number of images requested

Fragment-media quota consumption

API	Number of quota points consumed per request	Shared quota (N)
GetMediaForFragmentList	Number of fragments in the Fragments parameter	500 quota points per second, per stream [h]
GetClip	Number of fragments in the resulting clip
GetMP4MediaFragment	1
GetTSFragment	1
GetImages	Max number of images requested

For example, with a quota of 500 fragment media per second, the following call patterns for a particular stream are supported:

• 5 requests per second to GetClip with 100 fragments in each clip.

• 100 requests per second to GetClip with 5 fragments in each clip.

• 2 requests per second to GetClip with 100 fragments in each clip and 3 requests per second to GetMediaForFragmentList in each clip.

• 400 requests per second to GetMP4MediaFragment and 100 requests per second to GetTSFragment.

These quotas have an important implication regarding the number of HLS and MPEG-DASH sessions that can be supported per stream. There's no limit to the number of HLS and DASH sessions that can be in use by media players at a given time. Therefore, it's important that the playback application doesn't allow too many sessions to be in use concurrently. The following two examples describe how to determine the number of concurrent playback sessions that can be supported:

Example 1: Live streaming

In a live streaming scenario with HLS with 1 second duration fragments, an audio and video track, and MaxMediaPlaylistFragmentResults set to five, a media player typically makes two calls to GetHLSMediaPlaylist per second. One call is for the latest video metadata and another for the corresponding audio metadata. The two calls consume five fragment metadata quota points each. It also makes two calls to GetMP4MediaFragment per second: one call for the latest video and another for the corresponding audio. Each call consumes a single fragment media token, so two tokens are consumed in total.

In this scenario, up to 250 concurrent playback sessions can be supported. With 250 sessions, this scenario consumes 2,500 fragment metadata quota points per second (well below the 10,000 quota) and 500 fragment media quota points per second.

Example 2: On-demand playback

In an on-demand playback scenario of a past event with MPEG-DASH, an audio and video track and MaxManifestFragmentResults set to 1,000, a media player typically calls GetDASHManifest once at the start of the session (consuming 1,000 fragment metadata quota points) and it calls GetMP4MediaFragment at a rate of up to 5 times per second (consuming 5 fragment media quota points) until all fragments are loaded. In this scenario, up to 10 new sessions can be started per second (right at the 10,000 fragment metadata per second quota), and up to 100 sessions can be actively loading fragment media at a rate of 5 per second (right at the 500 fragment media per second quota).

You can use ArchivedFragmentsConsumed.Metadata and ArchivedFragmentsConsumed.Media to monitor the consumption of fragment metadata and fragment media quota points, respectively. For information about monitoring, see Monitoring Amazon Kinesis Video Streams.

Streaming metadata service quotas

The following service quotas apply to adding streaming metadata to a Kinesis video stream:

• You can prepend up to 10 metadata items to a fragment.

• A fragment metadata name can be up to 128 bytes in length.

• A fragment metadata value can be up to 256 bytes in length.

• A fragment metadata name can't begin with the string "AWS". If such a metadata item is added, the putFragmentMetadata method in the PIC returns a STATUS_INVALID_METADATA_NAME error (error code 0x52000077). Your application can then either ignore the error (the PIC doesn't add the metadata item), or respond to the error.

Producer SDK quotas

The following table contains the current quotas for values in the SDK. See Upload to Kinesis Video Streams for more information.

Note

Before setting these values, you must validate your inputs. The SDK doesn't validate these limits, and a runtime error occurs if the limits are exceeded.

Value	Limit	Notes
Max stream count	128	The maximum number of streams that a producer object can create. This is a soft limit (you can request an increase). It guarantees that the producer doesn't accidentally create streams recursively.
Max device name length	128 characters
Max tag count	50 per stream
Max stream name length	256 characters
Min storage size	10 MiB = 10 * 1024 * 1024 bytes
Max storage size	10 GiB = 10 * 1024 * 1024 * 1024 bytes
Max root directory path length	4,096 characters
Max auth info length	10,000 bytes
Max URI string length	10,000 characters
Max tag name length	128 characters
Max tag value length	1,024 characters
Min security token period	30 seconds
Security token grace period	40 minutes	If the specified duration is longer, it's limited to this value.
Retention period	0 or greater than one hour	0 indicates no retention.
Min cluster duration	1 second	The value is specified in 100 ns units, which is the SDK standard.
Max cluster duration	30 seconds	The value is specified in 100 ns units, which is the SDK standard. The backend API can enforce a shorter cluster duration.
Max fragment size	50 MB	For more information, see Amazon Kinesis Video Streams service quotas.
Max fragment duration	20 seconds	For more information, see Amazon Kinesis Video Streams service quotas.
Max connection duration	45 minutes	The backend closes the connection after this time. The SDK rotates the token and establishes a new connection within this time.
Max ACK segment length	1,024 characters	Maximum segment length of the acknowledgement sent to the ACK parser function.
Max content type string length	128 characters
Max codec ID string length	32 characters
Max track name string length	32 characters
Max codec private data length	1 MiB = 1 * 1024 * 1024 bytes
Min timecode scale value length	100 ns	The minimum timecode scale value to represent the frame timestamps in the resulting MKV cluster. The value is specified in increments of 100 ns, which is the SDK standard.
Max timecode scale value length	1 second	The maximum timecode scale value to represent the frame timestamps in the resulting MKV cluster. The value is specified in increments of 100 ns, which is the SDK standard.
Min content view item count	10
Min buffer duration	20 seconds	The value is specified in increments of 100 ns, which is the SDK standard.
Max update version length	128 characters
Max ARN length	1024 characters
Max fragment sequence length	128 characters
Max retention period	10 years