Spanner Data Boost is a fully managed, serverless service that providesindependent compute resources for supported Spanner workloads.Data Boost lets you execute analytics queries and data exportswith near-zero impact to existing workloads on the provisionedSpanner instance. The service consists of Spannerclusters that Google manages at the region level. For eligible queries thatrequest Data Boost, Spanner routes the workload tothese servers transparently. Eligible queries are those for which the firstoperator in the query execution plan is a distributed union. These queries don'thave to change to take advantage of Data Boost.

Data Boost is most impactful in the following scenarios where youwant to avoid negative impacts to the existing transactional system due toresource contention:

• Ad hoc or infrequent queries that involve processing large amounts of data.A typical example is afederatedqueryfrom BigQuery to Spanner.
• Reporting or data export jobs. An example is a Dataflow job toexport Spanner data to Cloud Storage.

The following diagram illustrates how Data Boost coordinates withthe Spanner instance to provide independent compute resources.

Benefits

Data Boost offers the following benefits:

• Provides workload isolation. You can run supported queries against thelatest data with near-zero impact on existing transactional workloadsregardless of query complexity or amount of data processed.
• Provides equal or better latency.
• Prevents over-provisioning of Spanner instances just tosupport occasional analytics queries.
• Offers a high degree of scalability with greater query parallelism thatscales elastically with burst loads.
• Provides comprehensive metrics, which let administrators identify the mostexpensive queries and determine the cost component to optimize.Administrators can then verify the impact of their optimizations bymonitoring the query's serverless processing unit consumption in its nextexecution.
• Requires no additional operational overhead. There is no extra service tomanage, no capacity planning or provisioning, no need to wait for scaling,and no maintenance.

Permissions

Any principal that runs a query or export that requests Data Boostmust have thespanner.databases.useDataBoost Identity and Access Management (IAM)permission. We recommend that you use theCloud Spanner Database Reader With DataBoost (roles/spanner.databaseReaderWithDataBoost)IAM role.

Billing and quotas

You pay only for actual processing units used by queries that run onData Boost. Administrators can set limits on usage to avoid costoverruns.

What's next

• Run federated queries with Data Boost
• Export data with Data Boost
• Use Data Boost in your applications
• Monitor Data Boost usage
• Monitor and manage Data Boost quota usage