Caching basics — Questions & Answers

Test your understanding of caching fundamentals for inference results, including cache keys, model versions, time-to-live (TTL), and differences between client-side and server-side caching. This easy quiz will help reinforce best practices and key concepts in caching strategies.

This quiz contains 15 questions. Below is a complete reference of all questions, answer choices, and correct answers. You can use this section to review after taking the interactive quiz above.

1. Question 1: Basic definition of caching

   What is the main purpose of caching inference results in an application?

   • To permanently delete obsolete data
   • To encrypt all data before transmission
   • To store previously computed outputs for faster future access
   • To sequence data requests alphabetically
   Show correct answer

   Correct answer: To store previously computed outputs for faster future access

2. Question 2: Cache key components

   Which of the following elements should typically be included in a cache key for model inference results?

   • Just the server IP address
   • Only the current date
   • Randomly generated numbers
   • Model name and version, input data, and user identifier
   Show correct answer

   Correct answer: Model name and version, input data, and user identifier

3. Question 3: Effect of model versioning

   If a cache key does not include the model version, what might happen when the model is updated?

   • The cache will automatically reset
   • Old results may be falsely returned for new model versions
   • Cache will stop storing any data
   • All computation results will double in speed
   Show correct answer

   Correct answer: Old results may be falsely returned for new model versions

4. Question 4: Understanding TTL

   What does TTL (Time To Live) refer to in caching for inference results?

   • The maximum duration a cached result is considered valid
   • The timestamp of the last server reboot
   • The size limit for cache entries
   • Total transfer latency for input data
   Show correct answer

   Correct answer: The maximum duration a cached result is considered valid

5. Question 5: Cache refresh mechanism

   When the TTL for a cached result expires, what typically happens?

   • It gets encrypted again
   • The cached entry is invalidated and recomputed if needed
   • The cache entry is silently ignored forever
   • It turns into permanent storage
   Show correct answer

   Correct answer: The cached entry is invalidated and recomputed if needed

6. Question 6: Client-side caching scenario

   If a web browser stores inference results locally, what type of caching is this?

   • Database replication
   • Server-based caching
   • Client-side caching
   • Global cache synchronization
   Show correct answer

   Correct answer: Client-side caching

7. Question 7: Server-side caching definition

   What describes server-side caching in the context of inference results?

   • Results are synced through USB drives
   • Cache is kept only on the user's personal device
   • Results are stored on the application server for all clients
   • Each device stores its own results
   Show correct answer

   Correct answer: Results are stored on the application server for all clients

8. Question 8: Unique cache keys per request

   Why is it important to ensure cache keys are unique for different requests?

   • To use more server memory
   • To prevent returning incorrect results from unrelated inputs
   • To increase TTL automatically
   • To guarantee higher network latency
   Show correct answer

   Correct answer: To prevent returning incorrect results from unrelated inputs

9. Question 9: Impact of omitting input data from cache keys

   If the input data is not part of a cache key, what issue can occur?

   • TTL will not function properly
   • The cache will never be accessed
   • Different inputs may incorrectly share the same cached result
   • The server will crash instantly
   Show correct answer

   Correct answer: Different inputs may incorrectly share the same cached result

10. Question 10: Appropriate TTL setting

    Which TTL value would be most appropriate for frequently changing inference models?

    • A TTL of 5 years
    • An unlimited TTL
    • A shorter TTL, such as 1-5 minutes
    • No TTL at all
    Show correct answer

    Correct answer: A shorter TTL, such as 1-5 minutes

11. Question 11: Cache hit versus miss

    What is a cache hit in the context of inference result caching?

    • When the user refreshes their browser
    • When a requested inference result is found in the cache and returned
    • When the cache is too full to store results
    • When two clients exchange cache entries
    Show correct answer

    Correct answer: When a requested inference result is found in the cache and returned

12. Question 12: Best practices for cache invalidation

    Which action is a best practice for invalidating cached inference results when a model is updated?

    • Shorten input data
    • Disable caching for all users
    • Change the model version included in the cache key
    • Increase the cache storage size
    Show correct answer

    Correct answer: Change the model version included in the cache key

13. Question 13: Reducing redundant computation

    How does proper caching of inference results help reduce redundant computations?

    • By always recalculating outputs every time
    • By serving duplicate requests from cached data instead of re-computing
    • By limiting the number of API requests
    • By randomly dropping requests
    Show correct answer

    Correct answer: By serving duplicate requests from cached data instead of re-computing

14. Question 14: Choosing between cache locations

    Which is an advantage of server-side caching over client-side caching for inference results?

    • Server-side caching allows results to be shared among multiple users
    • Server-side cache has no storage limitations
    • Server caches can only be used on mobile devices
    • Only the client can access cached data
    Show correct answer

    Correct answer: Server-side caching allows results to be shared among multiple users

15. Question 15: Risks of stale cache

    What is a potential risk of having an excessively long TTL on cached inference results?

    • Clients may receive outdated or incorrect results
    • Input data will be randomly altered
    • Network connections become unstable
    • All inference models will crash
    Show correct answer

    Correct answer: Clients may receive outdated or incorrect results