August 2021 Summaries

A sample application was created to compare the performance of leading serverless databases—DynamoDB, MongoDB (Atlas), Firestore, Cassandra (Datastax Astra), FaunaDB, and Redis (Upstash)—by measuring the latency of fetching the top 10 news articles in a common web use case using serverless functions. The data consists of 7,001 real news articles from the New York Times API, and latency was measured by executing a specific query while excluding database connection and network latency. Each database had custom configurations, such as indexing and regional deployment, to optimize performance measurements. The study found varying performances, with special notes on consistency and replication for FaunaDB, and performance nuances like cold connection overhead for Firestore and non-updatable primary keys for Cassandra. The project is ongoing, with plans for code refactoring and further improvements, and encourages community feedback via social media platforms.

Edge caching is a technology designed to reduce latency in globally distributed systems by caching REST responses at edge locations worldwide, similar to a CDN, achieving an average global latency of 5ms. It is particularly beneficial for web and mobile applications that can access databases directly without a backend, as well as for multi-region serverless architectures where functions like AWS Lambda can run in different regions to further lower latency. Edge computing, including Cloudflare workers and other edge functions, benefits from Redis with edge caching due to its low latency and lightweight nature. Although edge caching is available only for GET requests and incurs additional costs, it allows for controlling cache response durations with headers and complements global database replication efforts by providing fast data access while retaining consistency.