HeliosProxy vs ShardingSphere-Proxy

HeliosProxy is a self-hosted, programmable PostgreSQL data-plane for routing connections, handling failover, replaying traffic, and supporting database upgrades. It provides session, transaction, and statement pooling, with multiplexing up to 100:1. Routing can account for replica lag, workload type, and failing backends through circuit breaking. Cache options include local memory, Redis-backed storage, and semantic tiers, with invalidation when writes occur. Transaction Replay can move sessions to a new primary and optionally replay reads or transactions when conditions permit, although prepared statements, temporary tables, open cursors, advisory locks, and extended-protocol SET statements are not restored. A WebAssembly plugin runtime offers five hook points, Ed25519 verification, and OCI distribution. The admin console includes topology, plugin, chaos, shadow-execution, and traffic-replay panels. Authentication supports JWT, OAuth 2.0, LDAP, and API keys. HeliosProxy works with PostgreSQL 12+ and compatible databases, and deployment documentation covers Docker, Kubernetes, and standalone setups. It is open source under Apache 2.0 and costs nothing.

ShardingSphere-Proxy is a self-hosted database proxy that gives clients a database server interface over a binary protocol. Its listed client protocols are MySQL, PostgreSQL, openGauss, and Firebird. It distributes storage and computing across underlying databases through sharding, and provides distributed transactions with a hybrid XA and BASE engine. Read/write splitting balances reads using SQL semantics and awareness of database topology. Other capabilities include migration with sharding, federated queries, transparent encryption, and data masking.

You can deploy it from a binary distribution, as a Docker image, or with Helm. The binary requires Java JRE 8 or higher; Docker needs no additional dependency. For a MySQL backend, the Proxy ext-lib directory needs mysql-connector-j-8.4.0.jar. The project is free at free. Support for system databases and tables such as information_schema and pg_catalog is limited, and some graphical database clients may show errors.