Workload management using reservations

This page describes how to use slot reservations to manage your BigQuery workloads.

Slot reservations

In BigQuery, slots are allocated in pools called reservations. Reservations let you manage capacity and isolate workloads in ways that make sense for your organization. For example, you might create a reservation named prod for production workloads, and a separate reservation named test for testing, so that test jobs don't compete for resources with production jobs. Or, you might create reservations for different departments in your organization to allocate compute costs.

Despite its name, the capacity in a reservation is not necessarily reserved. When you use autoscaling reservations, the capacity is automatically scaled up and down based on the demand. In addition, idle slots can be shared across reservations.

Reservation assignments

To use allocated slots in a reservation, you must assign it to one or more projects, folders, or organizations. When a job in a project runs, it uses slots from its assigned reservation. Resources can inherit assignments from their parent in the Google Cloud resource hierarchy. If a project is not assigned to a reservation, it inherits the assignment of the parent folder or organization, if it exists.

Projects use the single most specific reservation in the resource hierarchy to which they are assigned. A folder assignment overrides an organization assignment, and a project assignment overrides a folder assignment.

If a project doesn't have an assigned or inherited reservation, then the job uses on-demand pricing. For more information about the resource hierarchy, see organizing BigQuery resources.

Resources can be assigned to None to represent an absence of an assignment. Projects that are assigned to None always use on-demand pricing. A common use case for None assignments is to assign an organization to a reservation and then use None to opt certain projects or folders out of that reservation. For more information, see Assign a project to None.

When you create an assignment, you specify the job type for that assignment:

• QUERY: Use this reservation for non-continuous query jobs, including SQL, DDL, DML, and BigQuery ML (built-in models) queries.

• BACKGROUND: Use this reservation when you choose to use your own reservation to run your BigQuery search index management jobs, BigQuery change data capture (CDC), or BigLake metadata caching background jobs. Also use this reservation when you replicate source databases to BigQuery with Datastream's background apply operations. BACKGROUND reservations are not available in the Standard edition.

• CONTINUOUS (Preview): Use this reservation for continuous query jobs. To use continuous queries, you must enroll in the feature preview.

• ML_EXTERNAL: Use this reservation for BigQuery ML CREATE MODEL queries that use services that are external to BigQuery. For more information, see Assign slots to BigQuery ML workloads. ML_EXTERNAL reservations are not available in the Standard edition.

• PIPELINE: Use this reservation for load and extract jobs.

  By default, load and extract jobs are free and use a shared pool of slots. BigQuery does not guarantee capacity availability for this shared pool or the throughput you see. If you are loading large amounts of data, your job might wait for slots to become available. In that case, you might want to purchase dedicated slots and assign PIPELINE jobs to them. As a best practice, create an additional, dedicated reservation with Ignore idle slots enabled. For more information about idle slots, see Idle slots.

  When load and extract jobs are assigned to a reservation, they lose access to the free pool. You should Monitor resource utilization and jobs to ensure your reservations have enough capacity to perform better than using the free pool.

You can't allocate individual slots to specific assignments. The BigQuery scheduler handles slot allocation for jobs using a reservation. For more information about how slots are used, see fair scheduling in BigQuery.

Combine reservations with on-demand billing

You can use capacity-based in one region and on-demand in another region. By default, all projects use on-demand billing. Within a region, you can opt a project, folder, or organization into capacity-based billing by assigning it to a reservation. For example, if you purchase a slot commitment in the US multi-region and assign your organization to the default reservation, your organization will be on capacity-based billing in the US multi-region, but will remain on on-demand billing in all other regions.

Within a region, you can combine capacity-based and on-demand billing by explicitly assigning projects to a reservation. Any project not assigned to a reservation remains on on-demand billing. You can also explicitly assign a project to use on-demand billing by assigning the reservation ID none. This is useful if you assign a folder or an organization to a reservation, but want some projects within that folder or organization to use on-demand billing. For more information, see assign a project to none.

Projects with on-demand billing use capacity that's separate from your committed capacity. Those projects don't affect the availability of your committed capacity.

Specifying an administration project

When you create commitments and reservations, they are associated with a Google Cloud project. This project manages the BigQuery reservations resources, and is the primary source of billing for these resources. This project does not have to be the same project that holds your BigQuery jobs or datasets.

As a best practice, create a dedicated project for reservation resources. This project is called the administration project, because it centralizes the billing and management of your commitments. Give this project a descriptive name like bq-COMPANY_NAME-admin. Then create one or more separate projects to hold your BigQuery jobs.

Only projects within the same Organization resource as the administration project can be assigned to a reservation. If the administration project is not part of an Organization, then only that project can use the slots.

The administration project is billed for the committed slots. Projects that use the slots are billed for storage but not billed for the slots. You can purchase more than one type of plan (for example, annual and three-year) and put the slots into the same administration project.

As a best practice, limit the number of administration projects. This helps to simplify billing management and slot allocation. One administration project for all of your organization's reservations is recommended where practical. Complex organizations might require additional administration projects to meet management or billing requirements.

Using multiple administration projects

In some cases, you might want to create more than one administration project:

• To separate costs from multiple reservations and commitments into different organizational units.
• To map one or more slot commitments to different sets of reservations.

Idle slot capacity is not shared between reservations in different administration projects.

When you're on the Capacity management page of the BigQuery Google Cloud console, you can view reservations and commitments only for the selected administrative project.

Sizing slot reservations

BigQuery is architected to scale linearly with increased resources. Depending on the workload, incremental capacity is likely to give you incremental performance. However, adding capacity also increases your costs. Therefore, choosing the optimal number of slots to purchase depends on your requirements for performance, throughput, and utility.

You can experiment with baseline and autoscaling slots to determine the best configuration of slots. For example, you can test your workload with 500 baseline slots, then 1,000, then 1,500, and 2,000, and observe the impact on performance.

After you allocate slots and run your workloads for at least seven days, you can use the slot estimator to analyze performance and model the effect of adding or reducing slots.

You can also examine the current slot usage of your projects, along with the chosen monthly price that you want to pay. On-demand workloads have a soft slot cap of 2,000 slots, but it is important to check how many slots are actually being used by your projects by using INFORMATION_SCHEMA.JOBS* views, Cloud Logging, the Jobs API, or BigQuery Audit logs. For more information, see Monitor reservations.

Manage workloads using reservations

You can use BigQuery reservations to allocate capacity between workloads, teams, or departments by creating additional reservations and assigning projects to these reservations. A reservation is an isolated pool of resources with the added benefit of being able to take advantage of idle capacity across your entire organization.

For example, you might have a total committed capacity of 1,000 slots and three workload types: data science, ELT, and BI.

• You can create a ds reservation with 500 slots, and assign all relevant Google Cloud projects to the ds reservation.
• You can create an elt reservation with 300 slots, and assign projects you use for ELT workloads to the elt reservation.
• You can create a bi reservation with 200 slots, and assign projects connected to your BI tools to the bi reservation.

Instead of partitioning your capacity across workload types, you might choose to create reservations for individual teams or departments.

Manage reservations in different regions

Reservations are regional resources. Slots purchased and reservations created in one region cannot be used in any other region. Project, folders, and organizations can be assigned to reservations in one region and run on-demand in another. To manage reservations in another region, you must change the region in the BigQuery Capacity Management page:

1. In the BigQuery console, click Reservations.
2. Click the Location picker and select a region where you want to manage reservations.
3. After a region is selected, you can purchase slots, create reservations, and assign a project to a reservation.

Manage reservations in complex organizations

Reservations are organization-scoped resources. When you create reservations, you can assign capacity to any project within the same Google Cloud organization. Most BigQuery users use a single administration project for their reservations and commitments. This administration project is associated with a Cloud Billing account, which is billed for the capacity.

However, if you have a complex organization with multiple divisions that manage their own bills, you might want to use multiple administration projects. Note that idle slots can only be shared across reservations created in the same administration project. You should be aware of the quotas and limits for reservations and administration projects.

If you use multiple Google Cloud organizations, you must create at least one administration project for each organization, and then manage reservations and commitments for each organization in its related administration project. You can't share capacity across organizations.

Slot commitments

A slot commitment is a purchase of slots for a specified period of time. You can purchase slots in 50 slot increments, up to your regional slot quota. Capacity commitments are optional, but can provide cost savings for steady-state workloads. There is no limit on the number of commitments that you can create. You are charged from the moment that your commitment purchase is successful. For current pricing information, see capacity commitment pricing.

• Annual commitment. You purchase a 365-day commitment. You can choose whether to renew or convert to a different type of commitment plan after 365 days.

• Three-year commitment. You purchase a three year commitment. You can choose whether to renew or convert to a different type of commitment plan after 3 years (1,095 days).

At the end of the commitment period, your commitment will renew based on the selected renewal plan.

You are billed monthly for annual or three-year commitment plans. However, your financial commitment is for the entire commitment period and can't be canceled on a monthly basis. Your usage is updated in the billing report daily, which you can view at any time.

Slot commitments are subject to capacity availability. When you attempt to purchase slot commitments, the success of your purchase is not guaranteed. However, once your commitment purchase is successful, the capacity is available until the commitment expires.

If you purchase slot commitments before creating a reservation, then a reservation named default is created automatically as a convenience. The default reservation has no special behavior. You can create additional reservations if needed or use the default reservation.

Renew commitments

You select a renewal plan when you purchase a commitment. You can change the renewal plan for a commitment at any time until it expires. The following renewal plans are available:

• None. After the end of the commitment period, the commitment is removed. Reservations are not affected.
• Annual. After the end of the commitment period, your commitment renews for another year.
• Three-Year. After the end of the commitment period, your commitment renews for another three years.

For information about purchasing and renewing commitments, see Create a capacity commitment.

For example, if you purchased an annual commitment at 6:00 PM on October 5, 2019, then you started being charged at that second. You can delete or renew the commitment after 6:00 PM on October 4, 2020, noting that 2020 is a leap year. You can change the renewal plans prior to 6:00 PM on October 4, 2020 as follows:

• If you choose to renew to an annual commitment, then at 6:00 PM on October 4, 2020, your commitment renews for another year.
• If you choose to renew to a three-year commitment, then at 6:00 PM on October 4, 2020, your commitment renews for three years.

Note: The renewal process can take up to roughly one hour after the commitment expires. For example, if a commitment expires at 6:00 PM on October 4, 2020, the renewed commitment record appears in the system between around 6:00 PM and 7:00 PM on October 4, 2020. No on-demand charges are applied within this data update period as the effective start time for the created commitment is 6:00 PM.

Commitment expiration

You can't delete a commitment once you create it. To delete an annual or three-year commitment, set its renewal plan to NONE. After the commitment expires, it is deleted automatically. To learn more about commitment expirations, see Commitment expiration.

If you accidentally purchase a commitment or made a mistake when you configured your commitment, contact Cloud Billing Support for help.

Reservation limitations

• Reservations in one organization can't be shared with other organizations.
• You must use separate reservations and separate administration projects for each organization.
• Each organization can have a maximum of 10 administration projects with active reservations in a single location.
• Idle capacity can't be shared between organizations or between different administration projects within a single organization.
• Commitments and reservations are regional resources. Commitments purchased in one region or multi-region can't be used for reservations in any other regions or multi-regions, even when the single region location is co-located with the multi-region location. For example, you can't use a commitment purchased in the EU multi-region for a reservation in europe-west4.
• Commitments and reservations can't be moved from one region or multi-region to another.
• Commitments purchased in one administration project can't be moved to a different administration project.
• Commitments purchased with one edition can't be used with reservations of another edition.
• Idle slots aren't shared between reservations of different editions.
• Autoscaled slots aren't shareable as they scale down when no longer required.