Database Reference
In-Depth Information
How Many Queries Can You Run?
BigQuery imposes two different types of limits on queries: quotas and rate
limits. The quotas are the easiest to understand; you can run up to 20,000
queries a day that process up to 100 TB of data. (Note that although these
numbers are subject to change, it is extremely unlikely that the quotas will
be reduced.) If you're running lots of huge queries and those amounts aren't
enough for you, don't worry—these caps (unlike some other quotas that
BigQuery exposes) can be raised —they are put in place to prevent people
from unintentionally running up a large bill quickly. That said, 100 TB of
data processed is $500 (at current prices), so you can still spend a lot of
money in a 24-hour period. If you want to run more queries, contact Google
Cloud Support. They will likely want some more information about how you
use BigQuery to make sure you get the most for your money and aren't
abusing the system.
Query rate limits are designed to encourage users to spread their queries
out throughout the day to smooth out load on the query clusters responsible
for executing queries. These clusters, although large, are not infinite. To
make sure that they remain responsive even when one customer performs
a lot of queries, BigQuery imposes rate limits on how many queries a single
customer can send at once. These rate limits are somewhat complex; they're
designed to let users run giant queries if they want but also to more strongly
limit queries that process a lot of data.
The simplest rate limit is the concurrent query rate limit. You can run up
to 20 queries at once. The other rate limit is a bit more confusing: You can
run 1 query of any size and up to 19 other queries that add up to 1 TB of
simultaneous processing. The 1 TB portion of the limit can be thought of as a
pipe with fixed capacity. When the pipe is full, you cannot start new queries.
When one query completes, it opens up space in the pipe that you can fill
with another query. The number of bytes processed by a query determines
the amount of space the query takes in the pipe.
If the pipe was the only mechanism used, you could never run queries that
processed more than 1 TB of data. To make sure customers can run larger
queries, any query that doesn't fit in the 1 TB pipe fills the “large” query slot.
This slot doesn't need to be taken up by an actual large query; if the pipe is
full with two 500 GB queries and you run another query that processes only
1 MB that will go in the “large” query slot. That said, the “large” query slot
