Edit

How to handle regional capacity constraints in Azure AI Search

This article helps you decide what to do when your preferred Azure AI Search region is unavailable due to capacity constraints. It also provides evaluation criteria for selecting an alternative region.

Capacity constraint options

When a preferred Azure region is unavailable due to capacity constraints, you have two options:

  • Deploy to an alternative region.
  • Retry deployment during off-peak hours.

Deploy to an alternative region. (Recommended path) Azure AI Search is available across many Azure regions with consistent APIs, SDKs, SLAs, and compliance certifications. For most workloads, the operational difference between regions within the same geography is negligible. See the following section, Criteria for selecting an alternative region, for a full evaluation framework.

Retrying the service during off-peak hours is also a viable consideration. Capacity constraints are sometimes temporary. Retrying deployment during low-traffic periods, such as nights or weekends in UTC, might succeed when peak-hour attempts fail. This option isn't guaranteed and isn't a substitute for evaluating an alternative region. If retries don't succeed within a reasonable window, proceed with an alternative region.

Retry during off-peak hours when:

  • The deployment timeline allows for a delay of several days.
  • The constraint is likely the result of a temporary regional surge.
  • You prefer no architectural changes at this time.

Criteria for selecting an alternative region

Evaluate the following criteria before selecting an alternative region:

  • Network latency
  • Compliance
  • Availability Zones
  • Feature and model availability
  • Data residency and sovereignty
  • Pricing

Network latency

If existing services, such as applications or databases, remain in a different region than the new Azure AI Search deployment, each API call between them incurs cross-region round-trip time (RTT). For US-to-US regional pairs, this time is typically 26–50ms. For most search workloads, latency under 50ms RTT isn't perceptible to end users. Co-locating the application and the search service in the same region eliminates cross-region RTT entirely. For current region-to-region latency measurements, see Azure network round-trip latency statistics.

Compliance

Azure compliance certifications - including FedRAMP High, HIPAA, CJIS, DoD IL2, IRS 1075, PCI DSS, and StateRAMP - are applied at the geography level, not the individual region level. All commercial Azure regions within the same geography carry an identical compliance posture.

For more information, see:

Availability Zones

Availability Zones (AZ) are physically separate datacenters within an Azure region. When a search service has two or more replicas in an AZ-supported region, Azure automatically distributes them across zones at no extra cost and with no configuration required.

AZ support is relevant when the workload requires a 99.99% query SLA or operates in a regulated industry with documented high-availability requirements. For dev/test environments or workloads protected by a multiregion disaster recovery strategy, AZ support is generally not a blocking requirement.

For the current list of regions that support Availability Zones for Azure AI Search, see Azure regions with Availability Zones.

Feature and model availability

Not all Azure AI Search features and AI models are available in every region. Before selecting an alternative region, verify availability for the specific capabilities your workload depends on, including any Azure OpenAI or Azure AI Foundry models used in retrieval-augmented generation (RAG), semantic ranking, or AI enrichment pipelines.

When selecting an alternative region for Azure AI Search, verify model availability in the same region. Co-locating Azure AI Search with Azure OpenAI or Azure AI Foundry in the same region eliminates cross-service latency and simplifies compliance and data residency requirements.

For more information, see:

Data residency and sovereignty

Azure replicates data for resiliency within the same geography. For all commercial Azure regions within the United States geography, data at rest stays within the United States. This replication satisfies data residency requirements common in federal, state, and regulated industry workloads.

For more information, see:

Pricing

Azure AI Search service pricing is uniform across all regions within the same geography. Selecting an alternative region within the same geography doesn't change the service tier pricing.

For current pricing, see Azure AI Search pricing.

Move or restore an Azure AI Search service

You can back up and restore indexes, skillsets, indexers, and synonym maps to any region by using official scripts and tooling. This process enables both migration to an alternative region and a return to the original region when capacity becomes available.

For more information, see:

Related content

  • Last updated on