Quotas and limits
This document lists the quotas and limits that apply to BigQuery.
Google Cloud uses quotas to help ensure fairness and reduce spikes in resource use and availability. A quota restricts how much of a Google Cloud resource your Google Cloud project can use. Quotas apply to a range of resource types, including hardware, software, and network components. For example, quotas can restrict the number of API calls to a service, the number of load balancers used concurrently by your project, or the number of projects that you can create. Quotas protect the community of Google Cloud users by preventing the overloading of services. Quotas also help you to manage your own Google Cloud resources.
The Cloud Quotas system does the following:
- Monitors your consumption of Google Cloud products and services
- Restricts your consumption of those resources
- Provides a means to request changes to the quota value
In most cases, when you attempt to consume more of a resource than its quota allows, the system blocks access to the resource, and the task that you're trying to perform fails.
Quotas generally apply at the Google Cloud project level. Your use of a resource in one project doesn't affect your available quota in another project. Within a Google Cloud project, quotas are shared across all applications and IP addresses.
There are also limits on BigQuery resources. These limits are unrelated to the quota system. Limits cannot be changed unless otherwise stated.
By default, BigQuery quotas and limits apply on a per-project basis. Quotas and limits that apply on a different basis are indicated as such; for example, the maximum number of columns per table , or the maximum number of concurrent API requests per user . Specific policies vary depending on resource availability, user profile, Service Usage history, and other factors, and are subject to change without notice.
Quota replenishment
Daily quotas are replenished at regular intervals throughout the day, reflecting their intent to guide rate limiting behaviors. Intermittent refresh is also done to avoid long disruptions when quota is exhausted. More quota is typically made available within minutes rather than globally replenished once daily.
Request a quota increase
To increase or decrease most quotas, use the Google Cloud console. For more information, see Request a higher quota .
For step-by-step guidance through the process of requesting a quota increase in Google Cloud console, click Guide me:
Cap quota usage
To learn how you can limit usage of a particular resource by creating a quota override, see Create quota override .
Required permissions
To view and update your BigQuery quotas in the Google Cloud console, you need the same permissions as for any Google Cloud quota. For more information, see Google Cloud quota permissions .
Troubleshoot
For information about troubleshooting errors related to quotas and limits, see Troubleshooting BigQuery quota errors .
Jobs
Quotas and limits apply to jobs that BigQuery runs on your behalf whether they are run by using Google Cloud console, the bq command-line tool, or programmatically using the REST API or client libraries.
Query jobs
The following quotas apply to query jobs created automatically by
running interactive queries, scheduled queries, and jobs submitted by using the jobs.query
and query-type jobs.insert
API methods:
Quota | Default | Notes |
---|---|---|
Query usage per day
|
Unlimited | There is no limit to the number of bytes that can be processed by
queries in a project. View quota in Google Cloud console |
Query usage per day per user
|
Unlimited | There is no limit to the number of bytes that a user's queries can
process each day. View quota in Google Cloud console |
Cloud SQL federated query cross-region bytes per day
|
1 TB | If the BigQuery query processing location
and the
Cloud SQL instance location are different, then your query is a
cross-region
query. Your project can run up to 1 TB in cross-region queries
per day. See Cloud SQL
federated queries
. View quota in Google Cloud console |
Cross-cloud transferred bytes per day
|
1 TB | You can transfer up to 1 TB of data per day from an Amazon S3 bucket or
from Azure Blob Storage. For more information, see Cross-cloud
transfer from Amazon S3
and Azure
. View quota in Google Cloud console |
The following limits apply to query jobs created automatically by
running interactive queries, scheduled queries, and jobs submitted by using the jobs.query
and query-type jobs.insert
API methods:
jobs.query
and query-type jobs.insert
API methods.A query or multi-statement query can execute for up to 6 hours, and then it fails. However, sometimes queries are retried. A query can be tried up to three times, and each attempt can run for up to 6 hours. As a result, it's possible for a query to have a total runtime of more than 6 hours.
CREATE MODEL
job timeout defaults to 24 hours, with the exception of time series, AutoML, and hyperparameter tuning jobs which timeout at 72 hours.
- Tables, views, UDFs, and table functions directly referenced by the query.
- Tables, views, UDFs, and table functions referenced by other views/UDFs/table functions referenced in the query.
The query is too large.
To stay within this
limit, consider replacing large arrays or lists with query parameters and breaking
a long query into multiple queries in the session.The query
is too large.
To stay within this limit, consider replacing large arrays or lists with
query parameters.The query is too
large.
To stay within this limit, consider replacing large arrays or lists with query
parameters.20,000 slots per organization
billingTierLimitExceeded
error.
For more information,
see billingTierLimitExceeded
.Although scheduled queries use features of the BigQuery Data Transfer Service , scheduled queries are not transfers, and are not subject to load job limits .
Export jobs
The following limits apply to jobs that export data
from BigQuery by using the bq command-line tool, Google Cloud console,
or the export-type jobs.insert
API method.
- Create a slot reservation
and specify
PIPELINE
as the assignment type . We will bill you using capacity-based pricing . - Use the
EXPORT DATA
SQL statement. We will bill you using either on-demand or capacity-based pricing , depending on how your project is configured. - Use the Storage Read API . We will bill you using the price for streaming reads . The expiration time is guaranteed to be at least 6 hours from session creation time.
- Create a slot reservation
and specify
PIPELINE
as the assignment type , or use an existingPIPELINE
assignment. We will bill you using capacity-based pricing . - Use the
EXPORT DATA
SQL statement. We will bill you using either on-demand or capacity-based pricing , depending on how your project is configured. - Use the Storage Read API . We will bill you using the price for streaming reads . The expiration time is guaranteed to be at least 6 hours from session creation time.
For more information about viewing your current export job usage, see View current quota usage .
Load jobs
The following limits apply when you load data
into BigQuery, using the
Google Cloud console, the bq command-line tool, or the load-type jobs.insert
API method.
Limit | Default | Notes |
---|---|---|
Load jobs per table per day
|
1,500 jobs | Load jobs, including failed load jobs, count toward the limit on the number of table operations per day for the destination table. For information about limits on the number of table operations per day for standard tables and partitioned tables, see Tables . |
Load jobs per day
|
100,000 jobs | Your project is replenished with a maximum of 100,000 load jobs quota every 24 hours. Failed load jobs count toward this limit. In some cases, it is possible to run more than 100,000 load jobs in 24 hours if a prior day's quota is not fully used. |
Maximum columns per table
|
10,000 columns | A table can have up to 10,000 columns. |
Maximum size per load job
|
15 TB | The total size for all of your CSV, JSON, Avro, Parquet, and ORC input files can be up to 15 TB. |
Maximum number of source URIs in job configuration
|
10,000 URIs | A job configuration can have up to 10,000 source URIs. |
Maximum number of files per load job
|
10,000,000 files | A load job can have up to 10 million total files, including all files matching all wildcard URIs. |
Maximum number of files in the source Cloud Storage bucket
|
Approximately 60,000,000 files | A load job can read from a Cloud Storage bucket containing up to approximately 60,000,000 files. |
Load job execution-time limit
|
6 hours | A load job fails if it executes for longer than six hours. |
Avro: Maximum size for file data blocks
|
16 MB | The size limit for Avro file data blocks is 16 MB. |
CSV: Maximum cell size
|
100 MB | CSV cells can be up to 100 MB in size. |
CSV: Maximum row size
|
100 MB | CSV rows can be up to 100 MB in size. |
CSV: Maximum file size - compressed
|
4 GB | The size limit for a compressed CSV file is 4 GB. |
CSV: Maximum file size - uncompressed
|
5 TB | The size limit for an uncompressed CSV file is 5 TB. |
Newline-delimited JSON (ndJSON): Maximum row size
|
100 MB | ndJSON rows can be up to 100 MB in size. |
ndJSON: Maximum file size - compressed
|
4 GB | The size limit for a compressed ndJSON file is 4 GB. |
ndJSON: Maximum file size - uncompressed
|
5 TB | The size limit for an uncompressed ndJSON file is 5 TB. |
If you regularly exceed the load job limits due to frequent updates, consider streaming data into BigQuery instead.
For information on viewing your current load job usage, see View current quota usage .
BigQuery Data Transfer Service load job quota considerations
Load jobs created by BigQuery Data Transfer Service transfers are included in
BigQuery's quotas on load jobs. It's important to consider
how many transfers you enable in each project to prevent transfers and other
load jobs from producing quotaExceeded
errors.
You can use the following equation to estimate how many load jobs are required by your transfers:
Number of daily jobs = Number of transfers x Number of tables x
Schedule frequency x Refresh window
Where:
-
Number of transfers
is the number of transfer configurations you enable in your project. -
Number of tables
is the number of tables created by each specific transfer type. The number of tables varies by transfer type:- Campaign Manager transfers create approximately 25 tables.
- Google Ads transfers create approximately 60 tables.
- Google Ad Manager transfers create approximately 40 tables.
- Google Play transfers create approximately 25 tables.
- Search Ads 360 transfers create approximately 50 tables.
- YouTube transfers create approximately 50 tables.
-
Schedule frequency
describes how often the transfer runs. Transfer run schedules are provided for each transfer type: -
Refresh window
is the number of days to include in the data transfer. If you enter 1, there is no daily backfill.
Copy jobs
The following limits apply to BigQuery jobs for copying tables
, including jobs
that create a copy, clone, or snapshot of a standard table, table clone, or
table snapshot.
The limits apply to jobs created by using the Google Cloud console, the
bq command-line tool, or the jobs.insert
method
that
specifies the copy
field
in the job configuration.
Copy jobs count toward these limits whether they succeed or fail.
Limit | Default | Notes |
---|---|---|
Copy jobs per destination table per day
|
See Table operations per day . | |
Copy jobs per day
|
100,000 jobs | Your project can run up to 100,000 copy jobs per day. |
Cross-region copy jobs per destination table per day
|
100 jobs | Your project can run up to 100 cross-region copy jobs for a destination table per day. |
Cross-region copy jobs per day
|
2,000 jobs | Your project can run up to 2,000 cross-region copy jobs per day. |
Number of source tables to copy
|
1,200 source tables | You can copy from up to 1,200 source tables per copy job. |
For information on viewing your current copy job usage, see Copy jobs - View current quota usage .
The following limits apply to copying datasets :
Limit | Default | Notes |
---|---|---|
Maximum number of tables in the source dataset
|
20,000 tables | A source dataset can have up to 20,000 tables. |
Maximum number of tables that can be copied per run to a destination
dataset in the same region
|
20,000 tables | Your project can copy 20,000 tables per run to a destination dataset that is in the same region. |
Maximum number of tables that can be copied per run to a destination
dataset in a different region
|
1,000 tables | Your project can copy 1,000 tables per run to a destination dataset that is in a different region. For example, if you configure a cross-region copy of a dataset with 8,000 tables in it, then BigQuery Data Transfer Service automatically creates eight runs in a sequential manner. The first run copies 1,000 tables. Twenty-four hours later, the second run copies 1,000 tables. This process continues until all tables in the dataset are copied, up to the maximum of 20,000 tables per dataset. |
Reservations
The following quotas apply to reservations :
View quotas in Google Cloud console
View quotas in Google Cloud console
us-east1
regionView quotas in Google Cloud console
-
asia-south1
-
asia-southeast1
-
europe-west2
-
us-central1
-
us-west1
View quotas in Google Cloud console
-
asia-east1
-
asia-northeast1
-
asia-northeast3
-
asia-southeast2
-
australia-southeast1
-
europe-north1
-
europe-west1
-
europe-west3
-
europe-west4
-
northamerica-northeast1
-
us-east4
-
southamerica-east1
View quotas in Google Cloud console
View quotas in Google Cloud console
View quotas in Google Cloud console
The following limits apply to reservations :
Limit | Value | Notes |
---|---|---|
5 projects per organization | The maximum number of projects within an organization that can contain a reservation or an active commitment for slots for a given location / region. | |
10 reservations per project | The maximum number of standard edition reservations per administration project within an organization for a given location / region. | |
200 reservations per project | The maximum number of Enterprise or Enterprise Plus edition reservations per administration project within an organization for a given location / region. | |
Maximum number of slots in a reservation that is associated with a
reservation assignment with a
CONTINUOUS
job type. |
500 slots | When you want to create a reservation assignment that has a CONTINUOUS
job type, the associated reservation can't have more than 500 slots. |
Datasets
The following limits apply to BigQuery datasets :
- Google Cloud console
- The bq command-line tool
- BigQuery client libraries
- The following API methods:
- The following DDL statements:
Tables
All tables
The following limits apply to all BigQuery tables.
Limit | Default | Notes |
---|---|---|
Maximum length of a column name
|
300 characters | Your column name can be at most 300 characters. |
Maximum length of a column description
|
1,024 characters | When you add a description to a column, the text can be at most 1,024 characters. |
Maximum depth of nested records
|
15 levels | Columns of type RECORD
can contain nested RECORD
types, also called child
records. The maximum nested depth limit is 15 levels.
This limit is independent of whether the records are scalar or
array-based (repeated). |
Standard tables
The following limits apply to BigQuery standard (built-in) tables :
Your project can make up to 1,500 table modifications per table per day, whether the modification appends data, updates data, or truncates the table. This limit cannot be changed and includes the combined total of all load jobs , copy jobs , and query jobs that append to or overwrite a destination table.
DML statements do not count toward the number of table modifications per day.
Streaming data does not count toward the number of table modifications per day.
- Google Cloud console
- The bq command-line tool
- BigQuery client libraries
- The following API methods:
- DDL statements on tables
DELETE
, INSERT
, MERGE
, TRUNCATE TABLE
, or UPDATE
statements to write
data to a table. Note that while DML statements count toward this limit, they are not subject to it if it is reached. DML operations have dedicated rate limits
. If you exceed this limit, you get an error message like Exceeded rate limits: too many table update operations for this
table
.
This error is transient; you can retry with an exponential backoff.
To identify the operations that count toward this limit, you can Inspect your logs . Refer to Troubleshoot quota errors for guidance on diagnosing and resolving this error.
External tables
The following limits apply to BigQuery tables with data stored on Cloud Storage in Parquet, ORC, Avro, CSV, or JSON format:
Limit | Default | Notes |
---|---|---|
Maximum number of source URIs per external table
|
10,000 URIs | Each external table can have up to 10,000 source URIs. |
Maximum number of files per external table
|
10,000,000 files | An external table can have up to 10 million files, including all files matching all wildcard URIs. |
Maximum size of stored data on Cloud Storage per external table
|
600 TB | An external table can have up to 600 terabytes across all input files. This limit applies to the file sizes as stored on Cloud Storage; this size is not the same as the size used in the query pricing formula. For externally partitioned tables, the limit is applied after partition pruning . |
Maximum number of files in the source Cloud Storage bucket
|
Approximately 60,000,000 files | An external table can reference a Cloud Storage bucket containing up to approximately 60,000,000 files. For externally partitioned tables, this limit is applied before partition pruning . |
Partitioned tables
The following limits apply to BigQuery partitioned tables .
Partition limits apply to the combined total of all load jobs , copy jobs , and query jobs that append to or overwrite a destination partition.
A single job can affect multiple partitions. For example, query jobs and load jobs can write to multiple partitions.
BigQuery uses the number of partitions affected by a job when determining how much of the limit the job consumes. Streaming inserts do not affect this limit.
For information about strategies to stay within the limits for partitioned tables, see Troubleshooting quota errors .
Your project can make up to 5,000 partition modifications per day, whether the modification appends data, updates data, or truncates an ingestion-time partitioned table.
DML statements do not count toward the number of partition modifications per day.
Your project can make up to 30,000 partition modifications per day for a column-partitioned table.
DML statements do not count toward the number of partition modifications per day.
Streaming data does not count toward the number of partition modifications per day.
- Google Cloud console
- The bq command-line tool
- BigQuery client libraries
- The following API methods:
- DDL statements on tables
DELETE
, INSERT
, MERGE
, TRUNCATE TABLE
, or UPDATE
statements to write
data to a table. If you exceed this limit, you get an error message like Exceeded rate limits: too many partitioned table update
operations for this table
.
This error is transient; you can retry with an exponential backoff.
To identify the operations that count toward this limit, you can Inspect your logs .
Table clones
The following limits apply to BigQuery table clones :
Limit | Default | Notes |
---|---|---|
Maximum number of clones and snapshots in a chain
|
3 table clones or snapshots | Clones and snapshots in combination are limited to a depth of 3. When you clone or snapshot a base table, you can clone or snapshot the result only two more times; attempting to clone or snapshot the result a third time results in an error. For example, you can create clone A of the base table, create snapshot B of clone A, and create clone C of snapshot B. To make additional duplicates of the third-level clone or snapshot, use a copy operation instead. |
Maximum number of clones and snapshots for a base table
|
1,000 table clones or snapshots | You can have no more than 1,000 existing clones and snapshots combined of a given base table. For example, if you have 600 snapshots and 400 clones, you reach the limit. |
Table snapshots
The following limits apply to BigQuery table snapshots :
Limit | Default | Notes |
---|---|---|
Maximum number of concurrent table snapshot jobs
|
100 jobs | Your project can run up to 100 concurrent table snapshot jobs. |
Maximum number of table snapshot jobs per day
|
50,000 jobs | Your project can run up to 50,000 table snapshot jobs per day. |
Maximum number of table snapshot jobs per table per day
|
50 jobs | Your project can run up to 50 table snapshot jobs per table per day. |
Maximum number of metadata updates per table snapshot per 10 seconds.
|
5 updates | Your project can update a table snapshot's metadata up to five times every 10 seconds. |
Maximum number of clones and snapshots in a chain
|
3 table clones or snapshots | Clones and snapshots in combination are limited to a depth of 3. When you clone or snapshot a base table, you can clone or snapshot the result only two more times; attempting to clone or snapshot the result a third time results in an error. For example, you can create clone A of the base table, create snapshot B of clone A, and create clone C of snapshot B. To make additional duplicates of the third-level clone or snapshot, use a copy operation instead. |
Maximum number of clones and snapshots for a base table
|
1,000 table clones or snapshots | You can have no more than 1,000 existing clones and snapshots combined of a given base table. For example, if you have 600 snapshots and 400 clones, you reach the limit. |
Views
The following quotas and limits apply to views and materialized views .
Logical views
The following limits apply to BigQuery standard views :
Limit | Default | Notes |
---|---|---|
Maximum number of nested view levels
|
16 levels | BigQuery supports up to 16 levels of nested views.
Creating views up to this limit is possible, but querying is limited to
15 levels. If the limit is exceeded, BigQuery returns an INVALID_INPUT
error. |
Maximum length of a GoogleSQL query used to define a view
|
256 K characters | A single GoogleSQL query that defines a view can be up to 256 K characters long. This limit applies to a single query and does not include the length of the views referenced in the query. |
Maximum number of authorized views per dataset
|
See Datasets . |
Materialized views
The following limits apply to BigQuery materialized views :
Limit | Default | Notes |
---|---|---|
Base table references (same dataset)
|
20 materialized views | Each base table can be referenced by up to 20 materialized views from the same dataset. |
Base table references (same project)
|
100 materialized views | Each base table can be referenced by up to 100 materialized views from the same project. |
Base table references (entire organization)
|
500 materialized views | Each base table can be referenced by up to 500 materialized views from the entire organization. |
Maximum number of authorized views per dataset
|
See Datasets . |
Search indexes
The following limits apply to BigQuery search indexes :
Limit | Default | Notes |
---|---|---|
Number of
CREATE INDEX
DDL statements per project per
region per day |
500 operations | Your project can issue up to 500 CREATE INDEX
DDL
operations every day within a region. |
Number of search index DDL statements per table per day
|
20 operations | Your project can issue up to 20 CREATE INDEX
or DROP INDEX
DDL operations per table per day. |
Maximum total size of table data per organization allowed for search
index creation that does not run in a reservation
|
100 TB in multi-regions; 20 TB in all other regions | You can create a search index for a table if the overall size of
tables with indexes in your organization is below your region’s limit:
100 TB for the US
and EU
multi-regions, and
20 TB for all other regions. If your index-management jobs run in your
own reservation
, then this limit doesn't apply. |
Vector indexes
The following limits apply to BigQuery vector indexes :
Limit | Default | Notes |
---|---|---|
Base table minimum number of rows
|
5,000 rows | A table must have at least 5,000 rows to create a vector index. |
Base table maximum number of rows
|
1,000,000,000 rows for index type IVF
200,000,000 for index type TREE_AH
|
A table can have at most 1,000,000,000 rows to create a IVF vector index, and 200,000,000 rows to create a TREE_AH vector index. |
Maximum size of the array in the indexed column
|
1,600 elements | The column to index can have at most 1,600 elements in the array. |
Minimum table size for vector index population
|
10 MB | If you create a vector index on a table that is under 10 MB, then the index is not populated. Similarly, if you delete data from a vector-indexed table such that the table size is under 10 MB, then the vector index is temporarily disabled. This happens regardless of whether you use your own reservation for your index-management jobs. Once a vector-indexed table's size again exceeds 10 MB, its index is populated automatically. |
Number of
CREATE VECTOR INDEX
DDL statements per project
per region per day |
500 operations | For each project, you can issue up to 500 CREATE VECTOR INDEX
operations per day for each region. |
Number of vector index DDL statements per table per day
|
10 operations | You can issue up to 10 CREATE VECTOR INDEX
or DROP VECTOR INDEX
operations per table per day. |
Maximum total size of table data per organization allowed for vector
index creation that does not run in a reservation
|
6 TB | You can create a vector index for a table if the total size of tables with indexes in your organization is under 6 TB. If your index-management jobs run in your own reservation , then this limit doesn't apply. |
Routines
The following quotas and limits apply to routines .
User-defined functions
The following limits apply to both temporary and persistent user-defined functions (UDFs) in GoogleSQL queries.
Limit | Default | Notes |
---|---|---|
Maximum output per row
|
5 MB | The maximum amount of data that your JavaScript UDF can output when processing a single row is approximately 5 MB. |
Maximum concurrent legacy SQL queries with Javascript UDFs
|
6 queries | Your project can have up to six concurrent legacy SQL queries that contain UDFs in JavaScript. This limit includes both interactive and batch queries. This limit does not apply to GoogleSQL queries. |
Maximum JavaScript UDF resources per query
|
50 resources | A query job can have up to 50 JavaScript UDF resources, such as inline code blobs or external files. |
Maximum size of inline code blob
|
32 KB | An inline code blob in a UDF can be up to 32 KB in size. |
Maximum size of each external code resource
|
1 MB | The maximum size of each JavaScript code resource is one MB. |
The following limits apply to persistent UDFs:
Limit | Default | Notes |
---|---|---|
Maximum length of a UDF name
|
256 characters | A UDF name can be up to 256 characters long. |
Maximum number of arguments
|
256 arguments | A UDF can have up to 256 arguments. |
Maximum length of an argument name
|
128 characters | A UDF argument name can be up to 128 characters long. |
Maximum depth of a UDF reference chain
|
16 references | A UDF reference chain can be up to 16 references deep. |
Maximum depth of a
STRUCT
type argument or output |
15 levels | A STRUCT
type UDF argument or output can be up to
15 levels deep. |
Maximum number of fields in
STRUCT
type arguments or output
per UDF |
1,024 fields | A UDF can have up to 1024 fields in STRUCT
type arguments
and output. |
Maximum number of JavaScript libraries in a
CREATE FUNCTION
statement |
50 libraries | A CREATE FUNCTION
statement can have up to 50 JavaScript
libraries. |
Maximum length of included JavaScript library paths
|
5,000 characters | The path for a JavaScript library included in a UDF can be up to 5,000 characters long. |
Maximum update rate per UDF per 10 seconds
|
5 updates | Your project can update a UDF up to five times every 10 seconds. |
Maximum number of authorized UDFs per dataset
|
See Datasets . |
Remote functions
The following limits apply to remote functions in BigQuery.
Limit | Default | Notes |
---|---|---|
Maximum number of concurrent queries that contain remote
functions
|
10 queries | You can run up to ten concurrent queries with remote functions per project. |
Maximum input size
|
5 MB | The maximum total size of all input arguments from a single row is 5 MB. |
HTTP response size limit (Cloud Run functions 1st gen)
|
10 MB | HTTP response body from your Cloud Run function 1st gen is up to 10 MB. Exceeding this value causes query failures. |
HTTP response size limit (Cloud Run functions 2nd gen or
Cloud Run)
|
15 MB | HTTP response body from your Cloud Run function 2nd gen or Cloud Run is up to 15 MB. Exceeding this value causes query failures. |
Max HTTP invocation time limit (Cloud Run functions 1st gen)
|
9 minutes | You can set your own time limit for your Cloud Run function 1st gen for an individual HTTP invocation, but the max time limit is 9 minutes . Exceeding the time limit set for your Cloud Run function 1st gen can cause HTTP invocation failures and query failure. |
HTTP invocation time limit (Cloud Run functions 2nd gen or
Cloud Run)
|
20 minutes | The time limit for an individual HTTP invocation to your Cloud Run function 2nd gen or Cloud Run. Exceeding this value can cause HTTP invocation failures and query failure. |
Maximum number of HTTP invocation retry attempts
|
20 | The maximum number of retry attempts for an individual HTTP invocation to your Cloud Run function 1st gen, 2nd gen, or Cloud Run. Exceeding this value can cause HTTP invocation failures and query failure. |
Table functions
The following limits apply to BigQuery table functions :
Limit | Default | Notes |
---|---|---|
Maximum length of a table function name
|
256 characters | The name of a table function can be up to 256 characters in length. |
Maximum length of an argument name
|
128 characters | The name of a table function argument can be up to 128 characters in length. |
Maximum number of arguments
|
256 arguments | A table function can have up to 256 arguments. |
Maximum depth of a table function reference chain
|
16 references | A table function reference chain can be up to 16 references deep. |
Maximum depth of argument or output of type
STRUCT
|
15 levels | A STRUCT
argument for a table function can be up to 15
levels deep. Similarly, a STRUCT
record in a table
function's output can be up to 15 levels deep. |
Maximum number of fields in argument or return table of type
STRUCT
per table function |
1,024 fields | A STRUCT
argument for
a table function can have up to 1,024 fields.
Similarly, a STRUCT
record in a table function's output can have up to 1,024 fields. |
Maximum number of columns in return table
|
1,024 columns | A table returned by a table function can have up to 1,024 columns. |
Maximum length of return table column names
|
128 characters | Column names in returned tables can be up to 128 characters long. |
Maximum number of updates per table function per 10 seconds
|
5 updates | Your project can update a table function up to five times every 10 seconds. |
Stored procedures for Apache Spark
The following limits apply for BigQuery stored procedures for Apache Spark :
You can use up to 2,400 CPUs for each location for each project, except in the following locations:
-
asia-south2
-
australia-southeast2
-
europe-central2
-
europe-west8
-
northamerica-northeast2
-
southamerica-west1
In these locations, you can use up to 500 CPUs for each location for each project.
If you run concurrent queries in a multi-region location and a single region location that is in the same geographic area, then your queries might consume the same concurrent CPU quota.
You can use up to 204.8 TB standard persistent disks for each location for each project. Queries that have already been processed don't consume this limit.
If you run concurrent queries in a multi-region location and a single region location that is in the same geographic area, then your queries might consume the same standard persistent disk quota.
Notebooks
All Dataform quotas and limits and Colab Enterprise quotas and limits apply to notebooks in BigQuery . The following limits also apply:
Limit | Default | Notes |
---|---|---|
Maximum notebook size
|
20 MB | A notebook's size is the total of its content, metadata, and encoding overhead. You can view the size of notebook content by expanding the notebook header, clicking View , and then clicking Notebook info . |
Maximum number of requests per second to Dataform
|
100 | Notebooks are created and managed through Dataform. Any action that creates or modifies a notebook counts against this quota. This quota is shared with saved queries. For example, if you make 50 changes to notebooks and 50 changes to saved queries within 1 second, you reach the quota. |
Saved queries
All Dataform quotas and limits apply to saved queries . The following limits also apply:
Limit | Default | Notes |
---|---|---|
Maximum saved query size
|
10 MB | |
Maximum number of requests per second to Dataform
|
100 | Saved queries are created and managed through Dataform. Any action that creates or modifies a saved query counts against this quota. This quota is shared with notebooks. For example, if you make 50 changes to notebooks and 50 changes to saved queries within 1 second, you reach the quota. |
Data manipulation language
The following limits apply for BigQuery data manipulation language (DML) statements:
Limit | Default | Notes |
---|---|---|
DML statements per day
|
Unlimited | The number of DML statements your project can run per day is unlimited. DML statements do not count toward the number of table modifications per day or the number of partitioned table modifications per day for partitioned tables. DML statements have the following limitations to be aware of. |
Concurrent
INSERT
DML statements per table per day |
1,500 statements | The first 1,500 INSERT
statements
run immediately after they are submitted. After this limit is reached,
the concurrency of INSERT
statements that write to a table
is limited to 10. Additional INSERT
statements are added to
a PENDING
queue. Up to 100 INSERT
statements
can be queued against a table at any given time. When an INSERT
statement completes, the next INSERT
statement is removed from the queue and run.If you must run DML INSERT
statements more frequently,
consider streaming data to your table using the Storage Write API
. |
Concurrent mutating DML statements per table
|
2 statements | BigQuery runs up to two concurrent mutating DML
statements ( UPDATE
, DELETE
, and MERGE
) for each table. Additional mutating DML statements
for a table are queued. |
Queued mutating DML statements per table
|
20 statements | A table can have up to 20 mutating DML statements in the queue waiting to run. If you submit additional mutating DML statements for the table, then those statements fail. |
Maximum time in queue for DML statement
|
6 hours | An interactive priority DML statement can wait in the queue for up to six hours. If the statement has not run after six hours, it fails. |
Maximum rate of DML statements for each table
|
25 statements every 10 seconds | Your project can run up to 25 DML statements every 10 seconds for each table. Both INSERT
and mutating DML statements contribute to this limit. |
For more information about mutating DML statements, see INSERT
DML concurrency
and UPDATE, DELETE, MERGE
DML concurrency
.
Multi-statement queries
The following limits apply to multi-statement queries in BigQuery.
Limit | Default | Notes |
---|---|---|
Maximum number of concurrent multi-statement queries
|
1,000 multi-statement queries | Your project can run up to 1,000 concurrent multi-statement queries . |
Cumulative time limit
|
24 hours | The cumulative time limit for a multi-statement query is 24 hours. |
Statement time limit
|
6 hours | The time limit for an individual statement within a multi-statement query is 6 hours. |
Recursive CTEs in queries
The following limits apply to recursive common table expressions (CTEs) in BigQuery.
Limit | Default | Notes |
---|---|---|
Iteration limit
|
500 iterations | The recursive CTE can execute this number of iterations. If this limit is exceeded, an error is produced. To work around iteration limits, see Troubleshoot iteration limit errors . |
Row-level security
The following limits apply for BigQuery row-level access policies :
Limit | Default | Notes |
---|---|---|
Maximum number of row-access policies per table
|
400 policies | A table can have up to 400 row-access policies. |
Maximum number of row-access policies per query
|
6000 policies | A query can access up to a total of 6000 row-access policies. |
Maximum number of
CREATE
/ DROP
DDL statements
per policy per 10 seconds |
5 statements | Your project can make up to five CREATE
or DROP
statements
per row-access policy resource every 10 seconds. |
DROP ALL ROW ACCESS POLICIES
statements per table per
10 seconds |
5 statements | Your project can make up to five DROP ALL ROW ACCESS POLICIES
statements per table every 10 seconds. |
Data policies
The following limits apply for column-level dynamic data masking :
Limit | Default | Notes |
---|---|---|
Maximum number of data policies per policy tag.
|
8 policies per policy tag | Up to eight data policies per policy tag. One of these policies can be used for column-level access controls . Duplicate masking expressions are not supported. |
BigQuery ML
The following limits apply to BigQuery ML.
Query jobs
All query job quotas and limits apply to GoogleSQL query jobs that use BigQuery ML statements and functions.
CREATE MODEL
statements
The following limits apply to CREATE MODEL
jobs:
Limit | Default | Notes |
---|---|---|
CREATE MODEL
statement queries per 48 hours for each project |
20,000 statement queries | Some models are trained by utilizing Vertex AI services , which have their own resource and quota management . |
Execution-time limit
|
24 hours or 72 hours | CREATE MODEL
job timeout defaults to 24 hours, with the exception of time series,
AutoML, and hyperparameter tuning jobs which timeout at 72
hours. |
Vertex AI and Cloud AI service functions
The following limits apply to functions that use Vertex AI large language models (LLMs) and Cloud AI services:
Function | Requests per minute | Rows per job | Number of concurrently running jobs |
---|---|---|---|
60 | 21,600 | 5 | |
200 | 72,000 | 5 | |
ML.GENERATE_TEXT
when using a remote model over the gemini-1.0-pro-vision
model in the us-central1
region |
100 | 20,000 | 1 |
ML.GENERATE_TEXT
when using a remote model over the gemini-1.0-pro-vision
model in regions other than us-central1
|
10 | 3,600 | 1 |
300 | 108,000 | 5 | |
ML.GENERATE_TEXT
when using a remote model over a gemini-1.0-pro
model in regions other than us-central1
|
10 | 3,600 | 5 |
ML.GENERATE_TEXT
when using a remote model over an Anthropic Claude model |
120 | 43,200 | 5 |
1,600 | 576,000 | 5 | |
300 | 108,000 | 5 | |
ML.GENERATE_EMBEDDING
when used with remote models over Vertex AI multimodalembedding
models in supported European single regions
|
120 | 14,000 | 5 |
ML.GENERATE_EMBEDDING
when used with remote models over Vertex AI multimodalembedding
models in regions other than supported European single regions
|
600 | 25,000 | 5 |
ML.GENERATE_EMBEDDING
when used with remote models over Vertex AI text-embedding
and text-multilingual-embedding
models
in the us-central1
region |
1,500 | 2,700,000 | 1 |
ML.GENERATE_EMBEDDING
when used with remote models over Vertex AI text-embedding
and text-multilingual-embedding
models
in regions other than the us-central1
|
100 | 180,000 | 1 |
600 | 216,000 | 5 | |
200 | 72,000 | 5 | |
1,800 | 648,000 | 5 | |
6,000 | 2,160,000 | 5 | |
600 | 21,600 | 5 |
For more information about quota for Vertex AI LLMs and the Cloud AI service APIs, see the following documents:
- Generative AI on Vertex AI quota limits
- Cloud Translation API quota and limits
- Vision API quota and limits
- Natural Language API quota and limits
- Document AI quota and limits
- Speech-to-Text quota and limits
Rows-per-job quota represents the highest theoretical number of rows that the
system can handle within a 6-hour timeframe. The actual number of processed
rows depends on many other factors, including input size and network condition.
For example, ML.TRANSCRIBE
can process more short audios than long audios.
To request more quota for the BigQuery ML functions, adjust the quota for the associated Vertex AI LLM or Cloud AI service first, and then send an email to bqml-feedback@google.com and include information about the adjusted LLM or Cloud AI service quota. For more information about how to request more quota for these services, see Request a higher quota .
Quota definitions
The following list describes the quotas that apply to Vertex AI and Cloud AI service functions:
- Functions that call a Vertex AI foundation model use one Vertex AI quota, which is queries per minute (QPM). In this context, the queries are request calls from the function to the Vertex AI model's API. The QPM quota applies to a base model and all versions, identifiers, and tuned versions of that model. For more information on the Vertex AI foundation model quotas, see Quotas per region and model .
- Functions that call a Cloud AI service use the target service's request quotas. Check the given Cloud AI service's quota reference for details.
-
BigQuery ML uses three quotas:
-
Requests per minute. This quota is the limit on the number of request calls per minute that functions can make to the Vertex AI model's or Cloud AI service's API. This limit applies to each project.
For functions that call a Vertex AI foundation model, the number of request calls per minute varies depending on the Vertex AI model endpoint, version, and region. This quota is conceptually the same as the QPM quota used by Vertex AI, but it might have a smaller value than the QPM quota for a corresponding model.
-
Rows per job. This quota is the limit on the number of rows allowed for each query job.
-
Number of concurrently running jobs. This quota is the limit per project on the number of SQL queries that can run at the same time for the given function.
-
The following examples show how to interpret quota limitations in typical situations:
-
I have a quota of 1,000 QPM in Vertex AI, so a query with 100,000 rows should take around 100 minutes. Why is the job running longer?
Job runtimes can vary even for the same input data. In Vertex AI, remote procedure calls (RPCs) have different priorities in order to avoid quota drainage. When there isn't enough quota, RPCs with lower priorities wait and possibly fail if it takes too long to process them.
-
How should I interpret the rows per job quota?
In BigQuery, a query can execute for up to six hours. The maximum supported rows is a function of this timeline and your Vertex AI QPM quota, in order to ensure that BigQuery can complete query processing in six hours. Since typically a query can't use the whole quota, this is a lower number than your QPM quota multiplied by 360.
-
What happens if I run a batch inference job on a table with more rows than the rows per job quota, for example 10,000,000 rows?
BigQuery only processes the number of rows specified by the rows per job quota. You are only charged for the successful API calls for that number of rows, instead of the full 10,000,000 rows in your table. For the rest of the rows, BigQuery responds to the request with a
A retryable error occurred: the maximum size quota per query has reached
error, which is returned in thestatus
column of the result. You can use this set of SQL scripts or this Dataform package to iterate through inference calls until all rows are successfully processed. -
I have many more rows to process than the rows per job quota. Will splitting my rows across multiple queries and running them simultaneously help?
No, because these queries are consuming the same BigQuery ML requests per minute quota and Vertex AI QPM quota. If there are multiple queries that all stay within the rows per job quota and number of concurrently running jobs quota, the cumulative processing exhausts the requests per minute quota.
BI Engine
The following limits apply to BigQuery BI Engine .
Limit | Default | Notes |
---|---|---|
250 GiB | Applies when using BI Engine with BigQuery. Applies in all cases except Looker Studio without native integration
. You can request an increase of the maximum reservation capacity for your projects. Reservation increases are available in most regions, and might take from 3 days to one week to process. |
|
100 GB | Applies when using BI Engine with Looker Studio without native integration . This limit does not affect the size of the tables that you query as BI Engine loads in-memory only the columns used in your queries, not the entire table. | |
10 GB | Applies when using BI Engine with Looker Studio without native integration . If you have a 100 GB reservation per project per location, BI Engine limits the reservation per table to 10 GB. The rest of the available reservation is used for other tables in the project. | |
500 partitions | Applies when using BI Engine with Looker Studio without native integration . BI Engine for Looker Studio supports up to a maximum of 500 partitions per table. | |
150 million | Applies when using BI Engine with Looker Studio without native integration . BI Engine for Looker Studio supports up to 150 million rows of queried data, depending on query complexity. |
Analytics Hub
The following limits apply to Analytics Hub :
Limit | Default | Notes |
---|---|---|
Maximum number of data exchanges per project
|
500 exchanges | You can create up to 500 data exchanges in a project. |
Maximum number of listings per data exchange
|
1,000 listings | You can create up to 1,000 listings in a data exchange. |
Maximum number of linked datasets per shared dataset
|
1,000 linked datasets | All Analytics Hub subscribers, combined, can have a maximum of 1,000 linked datasets per shared dataset. |
API quotas and limits
These quotas and limits apply to BigQuery API requests.
BigQuery API
The following quotas apply to BigQuery API (core) requests:
Quota | Default | Notes |
---|---|---|
Requests per day
|
Unlimited | Your project can make an unlimited number of BigQuery API requests per
day. View quota in Google Cloud console |
7.5 GB in multi-regions; 3.7 GB in all other regions | Your project can return a maximum of 7.5 GB of table row data per
minute via tabledata.list
in the us
and eu
multi-regions, and 3.7 GB of table row data per minute
in all other regions. This quota applies to the project that contains
the table being read. Other APIs including jobs.getQueryResults
and
fetching results from jobs.query
and jobs.insert
can also consume this quota.View quota in Google Cloud console The BigQuery Storage Read API
can sustain significantly higher throughput than |
The following limits apply to BigQuery API (core) requests:
Limit | Default | Notes |
---|---|---|
Maximum number of API requests per second per user per method
|
100 requests | A user can make up to 100 API requests per second to an API method. If a user makes more than 100 requests per second to a method, then throttling can occur. This limit does not apply to streaming inserts . |
Maximum number of concurrent API requests per user
|
300 requests | If a user makes more than 300 concurrent requests, throttling can occur. This limit does not apply to streaming inserts. |
Maximum request header size
|
16 KiB | Your BigQuery API request can be up to 16 KiB, including the request
URL and all headers. This limit does not apply to the request body, such
as in a POST
request. |
1,000 requests | Your project can make up to 1,000 jobs.get
requests per second. |
|
20 MB | By default, there is no maximum row count for the number of rows of
data returned by jobs.query
per page of results. However,
you are limited to the 20-MB maximum response size. You can alter the
number of rows to return by using the maxResults
parameter. |
|
20 MB | The maximum row size is approximate because the limit is based on the internal representation of row data. The limit is enforced during transcoding. | |
2 requests | Your project can make up to two projects.list
requests per second. |
|
1,000 requests | Your project can make up to 1,000 tabledata.list
requests per second. |
|
100,000 rows | A tabledata.list
call can return up to 100,000 table rows.
For more information, see Paging through results
using the API
. |
|
100 MB | The maximum row size is approximate because the limit is based on the internal representation of row data. The limit is enforced during transcoding. | |
10 requests | Your project can make up to 10 tables.insert
requests per second.
The tables.insert
method creates a new,
empty table in a dataset. The limit includes SQL statements that create
tables, such as CREATE TABLE
and queries that write results to destination tables
. |
BigQuery Connection API
The following quotas apply to BigQuery Connection API requests:
Quota | Default | Notes |
---|---|---|
Read requests per minute
|
1,000 requests per minute | Your project can make up to 1,000 requests per minute to
BigQuery Connection API methods that read connection data. View quota in Google Cloud console |
Write requests per minute
|
100 requests per minute | Your project can make up to 100 requests per minute to BigQuery Connection API
methods that create or update connections. View quota in Google Cloud console |
BigQuery Omni connections created per minute
|
10 connections created per minute | Your project can create up to 10 BigQuery Omni connections total across both AWS and Azure per minute. |
BigQuery Omni connection uses
|
100 connection uses per minute | Your project can use a BigQuery Omni connection up to 100 times per minute. This applies to operations which use your connection to access your AWS account, such as querying a table. |
BigQuery Migration API
The following limits apply to the BigQuery Migration API :
Limit | Default | Notes |
---|---|---|
Individual file size for batch SQL translation
|
10 MB | Each individual source and metadata file can be up to 10 MB.
This limit does not apply to the metadata zip file produced by the dwh-migration-dumper
command-line extraction tool. |
Total size of source files for batch SQL translation
|
1 GB | The total size of all input files uploaded to Cloud Storage can be up to 1 GB. This includes all source files, and all metadata files if you choose to include them. |
Input string size for interactive SQL translation
|
1 MB | The string that you enter for interactive SQL translation must not exceed 1 MB. When running interactive translations using the Translation API, this limit applies to the total size of all string inputs. |
Maximum configuration file size for interactive SQL translation
|
50 MB | Individual metadata files (compressed) and YAML config files in
Cloud Storage must not exceed 50 MB. If the file size exceeds 50 MB,
the interactive translator skips that configuration file during
translation and produces an error message. One method to reduce the
metadata file size is to use the —database
or –schema
flags to filter on databases when you generate the metadata
. |
The following quotas apply to the BigQuery Migration API . The following default values apply in most cases. The defaults for your project might be different:
Quota | Default | Notes |
---|---|---|
EDWMigration Service List Requests per minute EDWMigration Service List Requests per minute per user |
12,000 requests 2,500 requests |
Your project can make up to 12,000 Migration API List requests per minute. Each user can make up to 2,500 Migration API List requests per minute. View quotas in Google Cloud console |
EDWMigration Service Get Requests per minute EDWMigration Service Get Requests per minute per user |
25,000 requests 2,500 requests |
Your project can make up to 25,000 Migration API Get requests per minute. Each user can make up to 2,500 Migration API Get requests per minute. View quotas in Google Cloud console |
EDWMigration Service Other Requests per minute EDWMigration Service Other Requests per minute per user |
25 requests 5 requests |
Your project can make up to 25 other Migration API requests per minute. Each user can make up to 5 other Migration API requests per minute. View quotas in Google Cloud console |
Interactive SQL translation requests per minute Interactive SQL translation requests per minute per user |
200 requests 50 requests |
Your project can make up to 200 SQL translation service requests per minute. Each user can make up to 50 other SQL translation service requests per minute. View quotas in Google Cloud console |
BigQuery Reservation API
The following quotas apply to BigQuery Reservation API requests:
Quota | Default | Notes |
---|---|---|
Requests per minute per region
|
100 requests | Your project can make a total of up to 100 calls to BigQuery Reservation API
methods per minute per region. View quotas in Google Cloud console |
Number of
SearchAllAssignments
calls per minute per region |
100 requests | Your project can make up to 100 calls to the SearchAllAssignments
method per minute per region.View quotas in Google Cloud console |
Requests for
SearchAllAssignments
per minute per
region per user |
10 requests | Each user can make up to 10 calls to the SearchAllAssignments
method per minute per region.View quotas in Google Cloud console (In the Google Cloud console search results, search for per user.) |
BigQuery Data Policy API
The following limits apply for the Data Policy API ( preview ):
600 requests per minute per organization
600 requests per minute per organization
1800 requests per minute per organization
900 requests per minute per organization
This includes calls to:
IAM API
The following quotas apply when you use Identity and Access Management
features in BigQuery to retrieve and set IAM
policies, and to test IAM permissions. Data control language (DCL) statements
count towards SetIAMPolicy
quota.
Quota | Default | Notes |
---|---|---|
IamPolicy
requests per minute per user |
1,500 requests per minute per user | Each user can make up to 1,500 requests per minute per project. View quota in Google Cloud console |
IamPolicy
requests per minute per project |
3,000 requests per minute per project | Your project can make up to 3,000 requests per minute. View quota in Google Cloud console |
1,000 requests per minute per project | Your single-region project can make up to 1,000 requests per
minute. View quota in Google Cloud console |
|
2,000 requests per minute per project | Your multi-region project can make up to 2,000 requests per minute. View quota in Google Cloud console |
|
200 requests per minute per project | Your Omni-region project can make up to 200 requests per minute. View quota in Google Cloud console |
Storage Read API
The following quotas apply to BigQuery Storage Read API requests:
Quota | Default | Notes |
---|---|---|
Read data plane requests per minute per user
|
25,000 requests | Each user can make up to 25,000 ReadRows
calls per minute
per project.View quota in Google Cloud console |
Read control plane requests per minute per user
|
5,000 requests | Each user can make up to 5,000 Storage Read API metadata
operation calls per minute per project. The metadata calls include the CreateReadSession
and SplitReadStream
methods.View quota in Google Cloud console |
The following limits apply to BigQuery Storage Read API requests:
Limit | Default | Notes |
---|---|---|
Maximum row/filter length
|
1 MB | When you use the Storage Read API CreateReadSession
call, you are limited to a maximum length
of 1 MB for each row or filter. |
Maximum serialized data size
|
128 MB | When you use the Storage Read API ReadRows
call, the serialized representation of the data in an individual ReadRowsResponse
message cannot be larger than 128 MB. |
Maximum concurrent connections
|
2,000 in multi-regions; 400 in regions | You can open a maximum of 2,000 concurrent ReadRows
connections per project in the us
and eu
multi-regions, and 400 concurrent ReadRows
connections in
other regions. In some cases you may be limited to fewer concurrent
connections than this limit. |
Maximum per-stream memory usage
|
1.5 GB | The maximum per-stream memory is approximate because the limit is based on the internal representation of the row data. Streams utilizing more than 1.5 GB memory for a single row might fail. For more information, see Troubleshoot resources exceeded issues . |
Storage Write API
The following quotas apply to Storage Write API requests. The following quotas can be applied at the folder level. These quotas are then aggregated and shared across all child projects. To enable this configuration, contact Cloud Customer Care .
If you plan to request a higher quota limit , include the quota error message in your request to expedite processing.
Quota | Default | Notes |
---|---|---|
Concurrent connections
|
1,000 in a region; 10,000 in a multi-region | The concurrent connections quota is based on the client project that initiates the Storage Write API request, not the project containing the BigQuery dataset resource. The initiating project is the project associated with the API key or the service account . Your project can operate on 1,000 concurrent connections in
a region, or 10,000 concurrent connections in the When you use the default stream
in Java or Go, we recommend using Storage Write API multiplexing
to write to multiple destination tables with shared connections in order
to reduce the number of overall connections that are needed. If you are
using the Beam
connector with at-least-once semantics
, you can set UseStorageApiConnectionPool
to You can view usage quota and limits metrics for your projects in Cloud Monitoring
. Select the concurrent connections limit name based on your region. The options are |
Throughput
|
3 GB per second throughput in multi-regions; 300 MB per second in regions | You can stream up to 3 GBps in the us
and eu
multi-regions, and 300 MBps in other regions per project.View quota in Google Cloud console You can view usage quota and limits metrics for your projects in Cloud Monitoring
. Select the throughput limit name based on your region. The options are |
CreateWriteStream
requests |
10,000 streams every hour, per project per region | You can call CreateWriteStream
up to 10,000 times per hour
per project per region. Consider using the default stream
if you don't need exactly-once semantics.
This quota is per hour but the metric shown in the
Google Cloud console is per minute. |
Pending stream bytes
|
10 TB in multi-regions; 1 TB in regions | For every commit that you trigger, you can commit up to 10 TB in
the us
and eu
multi-regions, and
1 TB in other regions. There is no quota reporting on this quota. |
The following limits apply to Storage Write API requests:
Limit | Default | Notes |
---|---|---|
Batch commits
|
10,000 streams per table | You can commit up to 10,000 streams in each BatchCommitWriteStream
call. |
AppendRows
request size |
10 MB | The maximum request size is 10 MB. |
Streaming inserts
The following quotas and limits apply when you stream data into
BigQuery by using the legacy streaming API
.
For information about strategies to stay within these limits, see Troubleshooting quota errors
.
If you exceed these quotas, you get quotaExceeded
errors.
Limit | Default | Notes |
---|---|---|
Maximum bytes per second per project in the
us
and eu
multi-regions |
1 GB per second | Your project can stream up to 1 GB per second. This quota is cumulative within a given multi-region. In other words, the sum of bytes per second streamed to all tables for a given project within a multi-region is limited to 1 GB. Exceeding this limit causes If necessary, you can request a quota increase by contacting Cloud Customer Care . Request any increase as early as possible, at minimum two weeks before you need it. Quota increase takes time to become available, especially in the case of a significant increase. |
Maximum bytes per second per project in all other locations
|
300 MB per second | Your project can stream up to 300 MB per second in all locations
except the Exceeding this limit causes If necessary, you can request a quota increase by contacting Cloud Customer Care . Request any increase as early as possible, at minimum two weeks before you need it. Quota increase takes time to become available, especially in the case of a significant increase. |
Maximum row size
|
10 MB | Exceeding this value causes invalid
errors. |
HTTP request size limit
|
10 MB | Exceeding this value causes Internally the request is translated from HTTP JSON into an internal data structure. The translated data structure has its own enforced size limit. It's hard to predict the size of the resulting internal data structure, but if you keep your HTTP requests to 10 MB or less, the chance of hitting the internal limit is low. |
Maximum rows per request
|
50,000 rows | A maximum of 500 rows is recommended. Batching can increase performance and throughput to a point, but at the cost of per-request latency. Too few rows per request and the overhead of each request can make ingestion inefficient. Too many rows per request and the throughput can drop. Experiment with representative data (schema and data sizes) to determine the ideal batch size for your data. |
insertId
field length |
128 characters | Exceeding this value causes invalid
errors. |
For additional streaming quota, see Request a quota increase .