I'm working on a FastAPI application and need a way to search for a query parameter in a database table via an API endpoint. What's the best method to cache the data from this table? I'm currently designing a class with a timeout and last updated time field. When data is accessed, it will check if the timeout has been reached based on the current time, and if so, it will fetch fresh data from the database. I'm aware this could lead to latency when the timeout occurs, and I'm considering fetching the data asynchronously in the background. The database updates infrequently, about once per hour, and I'm thinking of setting the timeout to a minute. Is there a better approach?
5 Answers
The best caching strategy really depends on your specific requirements, such as how fast your queries need to be and how often your data changes. Consider your cache's location—whether it's in local RAM, SSD, or disk—and evaluate if your current database could be optimized for better performance. Additionally, databases like BigTable could provide the low latency you need without needing extensive setup changes.
Have you thought about using Redis or memcached? They're great for caching and can help reduce your data fetch times significantly. The only downside is that you'll need to manage these additional dependencies.
I considered it but I'm cautious about adding more complexity. Managing another process sounds overwhelming, even if loading data to Redis periodically could help.
Look into the CQRS pattern. It suggests caching read data and updating the cache whenever there’s an update. This could simplify your setup and improve efficiency.
Are you suggesting I should handle cached updates with a separate service or process?
If you're running multiple instances of your API, each with its own cache, you might end up serving stale data to some users. It could be beneficial to implement a centralized caching layer (like memcached) between your API instances and your database. This way, you can ensure consistency across all instances.
True, that's a good point!
Your current class implementation might not work as expected between requests since it’s stateless. Instead, consider using caching middleware that stores results either in memory or on disk. FastAPI has plugins that can help with caching as well. Also, regarding latency, be mindful that you won’t achieve sub-millisecond response times across the internet due to latency factors.
The latency I’m measuring is for data cached in RAM or potentially Redis.
I prefer caching in RAM. Switching databases isn't really an option for me, but it's good to know! My goal is sub-millisecond latencies, but a couple of milliseconds is manageable too. My query is a straightforward lookup, but the latency fluctuations are a real issue.