Account Usage¶

Examples: User login metrics¶

Average number of seconds between failed login attempts by user (month-to-date):

select user_name,
       count(*) as failed_logins,
       avg(seconds_between_login_attempts) as average_seconds_between_login_attempts
from (
      select user_name,
             timediff(seconds, event_timestamp, lead(event_timestamp)
                 over(partition by user_name order by event_timestamp)) as seconds_between_login_attempts
      from login_history
      where event_timestamp > date_trunc(month, current_date)
      and is_success = 'NO'
     )
group by 1
order by 3;

Copy

Failed logins by user (month-to-date):

select user_name,
       sum(iff(is_success = 'NO', 1, 0)) as failed_logins,
       count(*) as logins,
       sum(iff(is_success = 'NO', 1, 0)) / nullif(count(*), 0) as login_failure_rate
from login_history
where event_timestamp > date_trunc(month, current_date)
group by 1
order by 4 desc;

Copy

Failed logins by user and connecting client (month-to-date):

select reported_client_type,
       user_name,
       sum(iff(is_success = 'NO', 1, 0)) as failed_logins,
       count(*) as logins,
       sum(iff(is_success = 'NO', 1, 0)) / nullif(count(*), 0) as login_failure_rate
from login_history
where event_timestamp > date_trunc(month, current_date)
group by 1,2
order by 5 desc;

Copy

Examples: Warehouse performance¶

This query calculates virtual warehouse performance metrics such as throughput and latency for 15-minute time intervals over the course of one day.

In the code sample below, you can replace CURRENT_WAREHOUSE() with the name of a warehouse to calculate metrics for that warehouse. In addition, change the time_from and time_to dates in the WITH clause to specify the time period.

WITH
params AS (
SELECT
    CURRENT_WAREHOUSE() AS warehouse_name,
    '2021-11-01' AS time_from,
    '2021-11-02' AS time_to
),

jobs AS (
SELECT
    query_id,
    time_slice(start_time::timestamp_ntz, 15, 'minute','start') as interval_start,
    qh.warehouse_name,
    database_name,
    query_type,
    total_elapsed_time,
    compilation_time AS compilation_and_scheduling_time,
    (queued_provisioning_time + queued_repair_time + queued_overload_time) AS queued_time,
    transaction_blocked_time,
    execution_time
FROM snowflake.account_usage.query_history qh, params
WHERE
    qh.warehouse_name = params.warehouse_name
AND start_time >= params.time_from
AND start_time <= params.time_to
AND execution_status = 'SUCCESS'
AND query_type IN ('SELECT','UPDATE','INSERT','MERGE','DELETE')
),

interval_stats AS (
SELECT
    query_type,
    interval_start,
    COUNT(DISTINCT query_id) AS numjobs,
    MEDIAN(total_elapsed_time)/1000 AS p50_total_duration,
    (percentile_cont(0.95) within group (order by total_elapsed_time))/1000 AS p95_total_duration,
    SUM(total_elapsed_time)/1000 AS sum_total_duration,
    SUM(compilation_and_scheduling_time)/1000 AS sum_compilation_and_scheduling_time,
    SUM(queued_time)/1000 AS sum_queued_time,
    SUM(transaction_blocked_time)/1000 AS sum_transaction_blocked_time,
    SUM(execution_time)/1000 AS sum_execution_time,
    ROUND(sum_compilation_and_scheduling_time/sum_total_duration,2) AS compilation_and_scheduling_ratio,
    ROUND(sum_queued_time/sum_total_duration,2) AS queued_ratio,
    ROUND(sum_transaction_blocked_time/sum_total_duration,2) AS blocked_ratio,
    ROUND(sum_execution_time/sum_total_duration,2) AS execution_ratio,
    ROUND(sum_total_duration/numjobs,2) AS total_duration_perjob,
    ROUND(sum_compilation_and_scheduling_time/numjobs,2) AS compilation_and_scheduling_perjob,
    ROUND(sum_queued_time/numjobs,2) AS queued_perjob,
    ROUND(sum_transaction_blocked_time/numjobs,2) AS blocked_perjob,
    ROUND(sum_execution_time/numjobs,2) AS execution_perjob
FROM jobs
GROUP BY 1,2
ORDER BY 1,2
)
SELECT * FROM interval_stats;

Copy

Note

Analyze different statement types separately (e.g., SELECT statements independent of INSERT or DELETE or other statements).

• The NUMJOBS value represents the throughput for that time interval.

• The P50_TOTAL_DURATION (median) and P95_TOTAL_DURATION (peak) values represent latency.

• The SUM_TOTAL_DURATION is the sum of the SUM_<job_stage>_TIME values for the different job stages (COMPILATION_AND_SCHEDULING, QUEUED, BLOCKED, EXECUTION).

• Analyze the <job_stage>_RATIO values when the load (NUMJOBS) increases. Look for ratio changes or deviations from the average.

• If the QUEUED_RATIO is high, there might not be sufficient capacity in the warehouse. Add more clusters or increase the warehouse size.

Examples: Warehouse credit usage¶

Credits used by each warehouse in your account (month-to-date):

select warehouse_name,
       sum(credits_used) as total_credits_used
from warehouse_metering_history
where start_time >= date_trunc(month, current_date)
group by 1
order by 2 desc;

Copy

Credits used over time by each warehouse in your account (month-to-date):

select start_time::date as usage_date,
       warehouse_name,
       sum(credits_used) as total_credits_used
from warehouse_metering_history
where start_time >= date_trunc(month, current_date)
group by 1,2
order by 2,1;

Copy

Examples: Data storage usage¶

Billable terabytes stored in your account over time:

select date_trunc(month, usage_date) as usage_month
  , avg(storage_bytes + stage_bytes + failsafe_bytes) / power(1024, 4) as billable_tb
from storage_usage
group by 1
order by 1;

Copy

Examples: User query totals and execution times¶

Total jobs executed in your account (month-to-date):

select count(*) as number_of_jobs
from query_history
where start_time >= date_trunc(month, current_date);

Copy

Total jobs executed by each warehouse in your account (month-to-date):

select warehouse_name,
       count(*) as number_of_jobs
from query_history
where start_time >= date_trunc(month, current_date)
group by 1
order by 2 desc;

Copy

Average query execution time by user (month-to-date):

select user_name,
       avg(execution_time) as average_execution_time
from query_history
where start_time >= date_trunc(month, current_date)
group by 1
order by 2 desc;

Copy

Average query execution time by query type and warehouse size (month-to-date):

select query_type,
       warehouse_size,
       avg(execution_time) as average_execution_time
from query_history
where start_time >= date_trunc(month, current_date)
group by 1,2
order by 3 desc;

Copy

Examples: Obtain a query count for every login event¶

Join columns from LOGIN_HISTORY, QUERY_HISTORY, and SESSIONS to obtain a query count for each user login event.

Note

The SESSIONS view records information starting on July 20-21, 2020, therefore the query result will only contain overlapping information for each of the three views starting from this date.

select l.user_name,
       l.event_timestamp as login_time,
       l.client_ip,
       l.reported_client_type,
       l.first_authentication_factor,
       l.second_authentication_factor,
       count(q.query_id)
from snowflake.account_usage.login_history l
join snowflake.account_usage.sessions s on l.event_id = s.login_event_id
join snowflake.account_usage.query_history q on q.session_id = s.session_id
group by 1,2,3,4,5,6
order by l.user_name
;

Copy