Choosing a Platform

Choosing an Independent Platform for BookNest

The independent platforms differ less in price than in what they run underneath, and that decides what can go wrong later:

Independent Kafka 129 platforms compared
Platform Engine Licence of the engine Runs where Best for
Aiven Apache Kafka; Inkless fork Apache 2.0; Inkless module AGPL-3.0 Five clouds, BYOC Upstream Kafka, any cloud
Instaclustr Apache Kafka Apache 2.0 Three clouds, BYOC, on premises Support contracts, hybrid
Redpanda 64,862 Own C++ engine BSL 1.1, enterprise licence Serverless, BYOC, self-hosted Low latency, simple operations
StreamNative Pulsar, Ursa, Kafka fork Apache 2.0 (Pulsar); Ursa as a service Cloud, BYOC Pulsar shops, lakehouse streams
WarpStream Own agents on object storage Proprietary BYOC Cheap, latency-tolerant volume

Diskless designs (WarpStream, Ursa, Aiven's Inkless and the open-source, Apache 2.0 AutoMQ (https://github.com/AutoMQ/automq 10,901 ), a Kafka fork that keeps its log in S3) trade produce latency for much lower storage and cross-zone costs; they matter at terabytes a day, not at BookNest's megabytes.

For BookNest the choice follows from the earlier sections. Development and tests run on the Docker 514 cluster of Kafka 4.x in Docker, which does everything a broker can. A first production cluster goes to Aiven Startup in the cloud and region where the shop's backend runs: it is upstream Kafka 4.3, so every listing in this chapter works unchanged, as Aiven's Free Kafka Plan proved on the Free plan; it bundles Karapace 637 for the Avro 129 schemas of Schema Registry and managed Connect for the sinks of Kafka Connect; and at $200-265 a month it costs less than the hyperscaler options that run real Kafka. Redpanda would also serve BookNest's clients, but only with the classic group protocol and without share groups (Redpanda), so it is a candidate when latency or operational simplicity outweighs those features. Whatever the choice, keep plain Kafka clients and a Confluent-compatible registry API, so Choosing and Migrating's tools can move the topics later.