Ensure that Dynamic Quota is enabled for your Azure AI Services instances. Dynamic Quota (also known as Dynamic Throttling) is an AI Services feature that lets a deployment use more quota than its set Tokens Per Minute (TPM) limit if extra capacity is available. Dynamic Quota can be enabled at the instance level by setting the "dynamicThrottlingEnabled" property to true. When enabled, your deployment can handle higher traffic before receiving HTTP 429 (Too Many Requests) errors, allowing you to process more requests sooner.
This rule resolution is part of the Conformity Security & Compliance tool for Azure.
efficiency
Dynamic Quota allows an Azure AI Services model deployment to opportunistically use more capacity than its assigned quota when available. Enabling Dynamic Quota (Dynamic Throttling) can help protect against resource exhaustion attacks by making your service more resilient to sudden spikes in traffic, including malicious ones. Dynamic Throttling doesn't act as a firewall, but it acts as a shock absorber for your service's capacity. It allows your application to remain available to legitimate users by opportunistically handling a burst of requests, whether they are from a legitimate, high-traffic event or a malicious attack.
Audit
To determine if Dynamic Quota (Dynamic Throttling) is enabled for Microsoft Azure AI Services instances, perform the following operations:
Checking Azure AI Services instances for Dynamic Throttling configuration settings using the Azure Console (Azure Portal) is not currently supported.Remediation / Resolution
To enable Dynamic Quota (Dynamic Throttling) for your Azure AI Services instances, perform the following operations:
Enabling Dynamic Throttling for Azure AI Services instances using the Azure Console (Azure Portal) is not currently supported.References
- Azure Official Documentation
- Azure OpenAI in Azure AI Foundry Models quotas and limits
- Azure OpenAI Dynamic quota (Preview)
- Azure Command Line Interface (CLI) Documentation
- az account list
- az account set
- az cognitiveservices account list
- az cognitiveservices account show
- az rest