Caching reduces latency, lowers infrastructure costs, and improves reliability by making services less dependent on databases or third-party services.
Avoid adding a cache prematurely for young products with low traffic, as it introduces unnecessary complexity, potential bugs, and additional failure points.
Use a read-through cache backed by S3 to store static, rarely updated configuration data, ensuring service startup reliability even when dependencies are down.
Implement a cache-aside pattern using an in-memory LRU cache backed by DynamoDB and populated via Kafka CDC streams to reduce database load for frequently accessed data.
For smaller datasets, use Kafka's compact topics to maintain a complete, up-to-date copy of the data in-memory for each service instance.
Use a reverse proxy like Varnish Cache with a robust invalidation strategy to dramatically reduce response times for services with expensive computations like server-side rendering.
Follow a simple flowchart to select the appropriate caching strategy based on whether the data is for startup, dynamic retrieval, or stable HTTP responses.
The discussion covers HTTP header caching, custom invalidation logic, handling the "thundering herd" problem, and the choice of JVM for high-performance services.
Bitpanda
Vienna, Austria
28:12Josip Stuhli
45:48Denis Washington & Olli Salonen
30:36Markus Kett
46:24David Leitner
24:22Jürgen Pilz
30:27Thijs Feryn
50:06Jens Happe
22:32George Asafev
Jobs that call for the skills explored in this talk.
&why GmbH
Berlin, Germany
Wavestone
Canton of Courbevoie-2, France
