Database node
Runs the HTAP engine. An MVCC row heap takes transactional writes, the .zyr columnar format takes analytical scans, and background compaction moves committed rows between them.
HTAP databaseMesh
Zyron nodes peer over the wire and address each other's data directly. Any client can hit any node, and a single SELECT joins a local heap table to a remote publication or a shared lake, with no external gateway or middleware in front.
Topology
A deployment is a set of peers. Each node can host a database only, a lake reader only, both, or be embedded inside an application. Any client can hit any node, a single query reaches across the mesh, and lake-holding nodes share the same object-store backing.
Dashed peer edges carry publications, subscriptions, and foreign scans. Any client can hit any node, and the lake-holding nodes share one object store.
Node roles
Every node speaks the same wire protocol and joins the same mesh. What it hosts is a deployment choice, not a different product.
Runs the HTAP engine. An MVCC row heap takes transactional writes, the .zyr columnar format takes analytical scans, and background compaction moves committed rows between them.
HTAP databaseHosts ZyronLake, a shared-storage table format on an append-only transaction log, running on local disk, a shared filesystem, or object storage.
ZyronLakeBoth surfaces on one node. Heap and lake tables are first-class in the same SQL, joined by the same HybridScan operator.
Hosts neither surface and runs embedded in an application, peering over the same wire protocol as every other node.
Federation
Nodes address each other's data directly, over the same wire protocol clients use. A query that enters one node can read from the others without anything deployed in between.
Instances publish and subscribe to each other, and changes move between them with exactly-once wire-level push.
A single query joins local heap tables to remote publications or to a shared lake, with no external gateway or middleware in front.
One operator bridges the row heap and .zyr segments whether the segments live on the local node or across the mesh.
Heap and lake tables are first-class in the same SQL. There is no separate query layer.
~5 µs
QUIC PostgreSQL handshake
Peer links speak PostgreSQL wire protocol v3 over TCP and QUIC, the same protocol every client uses. This number comes from committed benchmark result files, like everything on the performance page.
Endpoints
The mesh does not stop at other databases. Dynamic endpoints put query results in front of browsers and applications directly.
A CREATE ENDPOINT statement declares a dynamic REST, WebSocket, or SSE endpoint. The endpoint surfaces SQL directly to browsers and apps.
Container hosting on the mesh substrate.
A QUERY / POST / GET / WS surface across every client language, replacing the REST + GraphQL split for apps talking to Zyron.
Operations
Peering across real infrastructure needs identity, routing, and observability. All of it ships in the node, and the wider model behind it is covered on the security page.
OAuth 2.0, AWS IAM STS, and Kubernetes SA tokens, with four secret managers as credential sources.
Routing runs over host lists with role hints and DNS SRV, with health checks and automatic failover.
Peer traffic runs over mTLS with SPKI pinning, on the same wire protocol clients connect with.
Labeled metric families per subscriber, per publication, and per TLS direction.
Prebuilt binaries for Linux and Windows, no external dependencies, one command to a running server.