Cost optimization strategies in Snowflake - Time & Space Complexity

Time Complexity: Cost optimization strategies

O(n)

Understanding Time Complexity

We want to understand how the cost of running Snowflake operations grows as we use more resources or run more queries.

How does the amount of work affect the cost and time spent?

Scenario Under Consideration

Analyze the time complexity of the following operation sequence.

-- Create a warehouse
CREATE WAREHOUSE my_wh WITH WAREHOUSE_SIZE = 'SMALL';

-- Run a query using the warehouse
SELECT * FROM large_table WHERE condition = 'value';

-- Suspend the warehouse to save cost
ALTER WAREHOUSE my_wh SUSPEND;

-- Resume warehouse when needed
ALTER WAREHOUSE my_wh RESUME;

This sequence shows creating a compute resource, running a query, and managing the warehouse state to optimize cost.

Identify Repeating Operations

Identify the API calls, resource provisioning, data transfers that repeat.

• Primary operation: Running queries on the warehouse.
• How many times: Each query run uses the warehouse resources once.
• Additional operations: Suspending and resuming the warehouse happen as needed to save cost.

How Execution Grows With Input

As the number of queries increases, the total compute time and cost grow roughly in direct proportion.

Input Size (n)	Approx. API Calls/Operations
10 queries	10 query executions + some suspend/resume calls
100 queries	100 query executions + some suspend/resume calls
1000 queries	1000 query executions + some suspend/resume calls

Pattern observation: The cost and operations grow linearly with the number of queries run.

Final Time Complexity

Time Complexity: O(n)

This means the total cost and operations increase directly in proportion to how many queries you run.

Common Mistake

[X] Wrong: "Running more queries won't increase cost if the warehouse is already running."

[OK] Correct: Each query uses compute resources, so more queries mean more compute time and higher cost, even if the warehouse stays running.

Interview Connect

Understanding how cost grows with usage helps you design efficient systems and explain trade-offs clearly in real work and interviews.

Self-Check

"What if we used auto-suspend and auto-resume features for the warehouse? How would the time complexity change?"

Practice

1. Which Snowflake feature helps reduce costs by automatically stopping a warehouse when it's not in use?

easy

A. Auto-suspend

B. Multi-cluster warehouse

C. Data sharing

D. Continuous data protection

Solution

1. Step 1: Understand warehouse billing

   Snowflake charges for compute resources while warehouses run, even if idle.

2. Step 2: Identify cost-saving feature

   Auto-suspend stops the warehouse after inactivity, preventing unnecessary charges.

3. Final Answer:

   Auto-suspend -> Option A

4. Quick Check:

   Auto-suspend = stops warehouse when idle [OK]

Hint: Auto-suspend pauses compute to save money when idle [OK]

Common Mistakes:

• Confusing auto-suspend with auto-resume
• Thinking multi-cluster reduces cost automatically
• Assuming data sharing affects compute cost

2. Which of the following is the correct syntax to set a warehouse to auto-suspend after 5 minutes in Snowflake?

easy

A. ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = '5 minutes';

B. ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = 5;

C. ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = 300;

D. ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = TRUE;

Solution

1. Step 1: Understand AUTO_SUSPEND unit

   Snowflake expects AUTO_SUSPEND value in seconds, so 5 minutes = 300 seconds.

2. Step 2: Check syntax correctness

   ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = 300; uses correct syntax and value (300 seconds). Others use wrong units or types.

3. Final Answer:

   ALTER WAREHOUSE my_wh SET AUTO_SUSPEND = 300; -> Option C

4. Quick Check:

   Auto-suspend time in seconds = 300 [OK]

Hint: Auto-suspend time is in seconds, not minutes [OK]

Common Mistakes:

• Using minutes instead of seconds
• Setting AUTO_SUSPEND to TRUE/FALSE
• Using string values instead of integers

3. Consider this Snowflake SQL snippet:

ALTER WAREHOUSE wh1 SET AUTO_SUSPEND = 600;
ALTER WAREHOUSE wh1 SET AUTO_RESUME = TRUE;

What is the effect on cost when no queries run for 15 minutes?

medium

A. Warehouse suspends after 10 minutes, no charges after that

B. Warehouse keeps running, charges continue for 15 minutes

C. Warehouse suspends immediately, no charges after query ends

D. Warehouse suspends after 15 minutes, charges continue until then

Solution

1. Step 1: Analyze AUTO_SUSPEND setting

   Warehouse suspends after 600 seconds (10 minutes) of inactivity.

2. Step 2: Understand AUTO_RESUME behavior

   Warehouse auto-resumes on new queries, but here no queries run for 15 minutes.

3. Step 3: Calculate cost impact

   Charges stop after suspension at 10 minutes; no charges for last 5 minutes idle.

4. Final Answer:

   Warehouse suspends after 10 minutes, no charges after that -> Option A

5. Quick Check:

   Auto-suspend 600s stops charges after 10 min idle [OK]

Hint: Auto-suspend time controls idle billing stop [OK]

Common Mistakes:

• Assuming warehouse suspends immediately after last query
• Thinking charges continue for full 15 minutes
• Confusing auto-resume with auto-suspend timing

4. You set a warehouse with AUTO_SUSPEND = 0 and AUTO_RESUME = TRUE. What issue might this cause?

medium

A. Warehouse auto-resume is disabled, queries fail

B. Warehouse never suspends, causing continuous charges

C. Warehouse suspends after default 10 minutes ignoring 0

D. Warehouse suspends immediately, causing query failures

Solution

1. Step 1: Understand AUTO_SUSPEND = 0 meaning

   Setting AUTO_SUSPEND to 0 disables suspension; warehouse runs continuously.

2. Step 2: Check AUTO_RESUME effect

   AUTO_RESUME TRUE allows auto start but does not suspend warehouse automatically.

3. Step 3: Identify cost impact

   Warehouse never suspends, so charges accrue continuously even if idle.

4. Final Answer:

   Warehouse never suspends, causing continuous charges -> Option B

5. Quick Check:

   AUTO_SUSPEND=0 means no suspend, costs continue [OK]

Hint: AUTO_SUSPEND=0 disables suspend, causing constant charges [OK]

Common Mistakes:

• Thinking 0 means immediate suspend
• Assuming AUTO_RESUME disables charges
• Believing default suspend time applies

5. A company wants to minimize Snowflake costs by scheduling warehouse usage only during business hours (9am-5pm). Which combination of strategies best achieves this?

hard

A. Set warehouse to always on and rely on auto-resume to start queries

B. Use multi-cluster warehouse to handle peak loads during business hours

C. Disable auto-suspend and manually suspend warehouse after hours

D. Use auto-suspend with 1-minute timeout and schedule warehouse start/stop tasks at 9am and 5pm

Solution

1. Step 1: Understand scheduling needs

   Warehouse should run only during business hours to save costs outside those times.

2. Step 2: Evaluate options

   Use auto-suspend with 1-minute timeout and schedule warehouse start/stop tasks at 9am and 5pm combines auto-suspend (quick suspend) and scheduled start/stop tasks to control usage precisely.

3. Step 3: Rule out other options

   Set warehouse to always on and rely on auto-resume to start queries wastes cost by always on; C is manual and error-prone; D increases cost for peak handling.

4. Final Answer:

   Use auto-suspend with 1-minute timeout and schedule warehouse start/stop tasks at 9am and 5pm -> Option D

5. Quick Check:

   Scheduled start/stop + auto-suspend = cost control [OK]

Hint: Schedule start/stop plus auto-suspend for tight cost control [OK]

Common Mistakes:

• Leaving warehouse always on wastes money
• Relying on manual suspend risks forgetting
• Using multi-cluster increases cost, not reduces