DocsICP FilingConsole
官方文档
首页

Concurrency and QPS

更新时间：
复制 MD 格式
产品详情
我的收藏

This document explains the concepts of concurrency and Queries Per Second (QPS) and shows you how to monitor service usage, such as call volume and concurrency.

Basic concepts

  • Concurrency

    Concurrency is the number of requests a system can process at the same time. This concept applies to services in Intelligent Speech Interaction, such as Short Speech Recognition, Real-time Speech Recognition, Speech Transcriber - Express Edition, Speech Synthesis, and real-time long-form speech synthesis. In Intelligent Speech Interaction, concurrency limits are not shared between different services. However, for a single service, the concurrency limit is shared across all its projects (each identified by a different AppKey).

    For example, with Real-time Speech Recognition, each request you initiate (for example, by calling the start API) establishes a WebSocket connection. This request occupies one concurrent connection until it ends (for example, when you call the stop or close API). If your concurrency limit is 200, you can have a maximum of 200 active real-time transcription requests. If you exceed this limit, the service returns error code 40000005.

  • QPS

    QPS (Queries Per Second) is the number of requests processed per second. The QPS concept applies to offline services in Intelligent Speech Interaction, such as Speech Transcriber and Speech Transcriber - Off-peak Edition.

    For example, with Speech Transcriber (including the Off-peak Edition), the QPS limit defines the maximum number of API requests you can make per second. If you exceed the QPS limit, you may receive an error such as Throttling.User: Request was denied due to user flow control. We recommend adjusting the call frequency for your task submission and task query APIs.

Concurrency and QPS limits

Service

Free trial

Commercial

Short Speech Recognition

2 concurrent connections

200 concurrent connections

Real-time Speech Recognition

2 concurrent connections

200 concurrent connections

Speech Transcriber

Transcription audio duration: 2 hours of audio/day

Note

During the 3-month trial period for new users, you can transcribe 2 hours of audio for free every 24 hours. After you use the free quota, it replenishes after 24 hours.

Transcription audio duration: Unlimited

  • The user-level QPS limit for transcription requests using the POST method is 200.

  • The user-level QPS limit for transcription requests using the GET method is 500.

  • For the transcription result query API, the QPS limit is 1 for the same Taskid.

Speech Transcriber - Off-peak Edition

Trial not supported

Transcription audio duration: Unlimited

  • The user-level QPS limit for transcription requests using the POST method is 200.

  • The user-level QPS limit for transcription requests using the GET method is 500.

  • For the transcription result query API, the QPS limit is 1 for the same Taskid.

Speech Transcriber - Express Edition

Trial not supported

10 concurrent connections

Streaming Speech Synthesis (CosyVoice Model)

Trial not supported

10 concurrent connections

Speech Synthesis

2 concurrent connections

200 concurrent connections

real-time long-form speech synthesis

Trial not supported

100 concurrent connections

asynchronous long-form speech synthesis

Trial not supported

100 concurrent connections

Speaker Recognition

2 concurrent connections

200 concurrent connections

Sound Event Detection

2 concurrent connections

200 concurrent connections

Language Identification

2 concurrent connections

200 concurrent connections

Gender Recognition

2 concurrent connections

200 concurrent connections

Monitor concurrency and QPS

  1. Log in to the Intelligent Speech Interaction console and choose Monitoring and Statistics in the left-side navigation pane.

  2. At the top of the Monitoring and Statistics page, switch between the Speech Recognition, Speech Synthesis, and Speech Analysis tabs to view usage for each service.

  3. Select the service, project, and region, choose a time range, and then click Query.

    The page then displays usage data for call volume and QPS/concurrency for the selected service.

Upgrade concurrency

If the default concurrency limit is insufficient for your business needs after upgrading to the commercial version, follow these steps to increase it.

This section uses Real-time Speech Recognition as an example.

  1. Log in to the Intelligent Speech Interaction console and choose Service Management and Activation in the left-side navigation pane.

  2. On the Speech Recognition tab, click Upgrade Concurrency to the right of Real-time Speech Recognition.

    Note

    If the Upgrade Concurrency option is not visible, self-service upgrades are not currently supported. Please contact customer service for a manual assessment. You can upgrade your concurrency once the assessment is approved.

  3. On the purchase page, select the type, specifications, quantity, and subscription duration, and then click Buy Now.

    Note

    Usage data appears the day after purchase.

Related documentation

For more information about concurrency billing, see Billing Method.

Previous：BillingNext：Refund policy