You need a highly scalable, lowest-latency API that returns real-time stock data from an internal DB. Which architecture minimizes latency while scaling?

The correct option minimizes latency and scales by leveraging Memorystore for Redis as a cache. Redis is an in-memory data store, providing extremely low-latency access for frequently requested stock data. By querying the cache first, most requests are served very quickly. If data is not in the cache, the system falls back to the internal database, ensuring data availability while still benefiting from caching for common queries. This hybrid approach balances performance and data freshness. Storing all data in BigQuery is for analytical workloads, not real-time, low-latency APIs. Storing all data in Memorystore for Redis might exceed memory limits for a large dataset and doesn't inherently handle persistence for all data. Using Bigtable with Pub/Sub introduces queuing latency and Bigtable is optimized for high-throughput, low-latency reads but not necessarily the absolute lowest latency for all data like an in-memory cache.