AlgoX2 has its own native protocol with shared-memory delivery for maximum performance. It also natively speaks Kafka, NATS, Redis, HTTP, and WebSocket — all connecting to the same stream layer. Publish via one protocol, subscribe via another. No bridge services. No translation layers.

Need a protocol we don't list? AlgoX2 can describe any binary protocol once using a compact schema, then automatically generate fast, safe C++ code for parsing and serialization. Legacy protocols, proprietary formats, custom wire formats — we can add support quickly without hand-written adapters.

Full Kafka API

AlgoX2 supports the complete Kafka API — producers, consumers, Connect, Streams, Admin, and Transactions. Your existing clients, connectors, and tooling work without code changes. Point them at AlgoX2, evaluate side-by-side, and switch back at any time.

Why AlgoX2 is architecturally different

Kafka bundles ingress, ordering, replication, and storage into a single broker — creating shared bottlenecks under load. AlgoX2 separates these into three independent layers: Gateways (protocol ingress/egress), Sequencers (ordering and replication), and Disk processes (commit and journals). Each scales independently.

This separation eliminates the problems that force Kafka deployments to hundreds of servers:

• Connection explosion — Kafka brokers maintain TCP connections for every producer; context switches dominate CPU at scale. AlgoX2 gateways handle connections separately from sequencing.
• File explosion — Each Kafka topic-partition-replica creates multiple log segments, driving disks toward random I/O. AlgoX2 keeps a constant file count regardless of stream count — always sequential writes.
• Redundant replication — Kafka resends the same bytes over TCP to each follower. AlgoX2 uses internal multicast — the network switch replicates in one step.
• Fan-out bottleneck — Kafka's single partition leader serves all consumers. AlgoX2 spreads reads across all nodes — full cluster bandwidth to every consumer.