Google Cloud Pub/Sub vs Raystack Stencil

Google Cloud Pub/Sub is a cloud messaging service for asynchronous communication between services that produce messages and those that process them. It is intended for service-to-service use, including streaming analytics, data integration, middleware, and parallel task queues, rather than direct communication with end users or IoT clients. Communication typically has latency around 100 milliseconds. Synchronous cross-zone replication and per-message receipt tracking provide at-least-once delivery; Dataflow supports exactly-once processing of Pub/Sub streams. There are no shards or partitions, and provisioning scales automatically after quota is set. Message controls include seek and replay, dead letter topics, and attribute-based filtering. Pub/Sub offers fine-grained access controls, end-to-end encryption, and HIPAA compliance. It supports REST and gRPC APIs and client libraries for seven languages; API keys are not supported for authentication. The maximum message size is 10 MB. The free monthly tier is free for the first 10 GB of data per month. Pay-as-you-go is $40 a month beyond 10 GB, with throughput listed at $40 per TB and storage at $0.27 per GiB-month.

Raystack Stencil is a schema registry for managing schema definitions and checking data compatibility across applications. It supports Protobuf, Avro, and JSON, and keeps versioned histories of proto descriptor files organized by namespace and name. Uploads receive backward-compatibility checks by default, though some checks can be skipped. The service provides APIs for retrieving the latest version number for a schema name and namespace and downloading the proto descriptors contained in a specified message. Its server offers gRPC and HTTP interfaces, while a CLI can create, view, and search schemas. Listed client libraries cover Java, Go, JavaScript, and Clojure. Clients can download descriptor sets, deserialize Protobuf messages, look up descriptors, and refresh schema definitions with built-in strategies. Stencil can run from a Docker image or be compiled from source, with installation options for macOS, Windows, Linux, OpenBSD, and FreeBSD. The server requires PostgreSQL 13. Refreshing a descriptor after encountering unknown fields is supported only by the Java and Go clients. The project is open source under Apache-2.0.