DaaS-IoT routing connects logical identities. The sender specifies the destination DIN; the overlay determines the node, driver, and endpoint for the next hop.
This routing layer operates above the physical networks. IP, BLE, or a serial bus transfers the frame between adjacent endpoints; DaaS-IoT preserves the end-to-end destination and builds the path between nodes.
Identity-based routing
In a traditional network, an address often identifies both the endpoint and its location. In DaaS-IoT, these concepts are separate:
DIN destination identity route sequence of overlay nodes link + URI next physical transfer
A node may change its IP address, interface, or access technology. The DIN remains stable, and the mapping between identity and channel is updated.
This separation makes it possible to cross multiprotocol gateways:
Each hop manages the transfer to the next one. The DME origin and destination remain stable along the entire path.
Local view and forwarding
Each node maintains a table of known peers and available channels. Three cases may occur for a destination:
Direct channel
The node has an active mapping to the destination and delivers the DME to the associated driver.
Known route
The route associates the destination with a next hop. The DME is delegated to the intermediate node, which repeats the decision using its own view of the network.
Unknown destination
If no usable mapping exists, the node starts a bounded route search or returns an error, according to the configured policy.
Each node performs forwarding using its own local view of the graph.
Route discovery
A route search propagates a request among nodes that are able to forward it. Two limits prevent the search from continuing indefinitely:
- a hop limit defines how many intermediate nodes it may cross;
- a timeout defines how long the requester waits for a response.
The destination, or a node that has a valid route to it, produces the response. The collected information is added to local knowledge and can be reused for subsequent DMEs.
A low hop limit contains traffic and restricts the explored portion of the topology. A higher value expands the search and its cost. The parameter must be sized according to the expected topology depth.
Route cache and expiration
A route that is valid at one point may become incorrect after a disconnection or reconfiguration. Routes therefore have a limited lifetime.
Expiration balances stability and adaptation:
- short-lived routes react quickly to changes and generate more discovery traffic;
- long-lived routes reduce control traffic and increase the time stale paths remain in use.
A next-hop failure invalidates the path before its natural expiration. If an alternative exists, the core can search for it and delegate the DME again.
Auto-routing after a failure
Route discovery may be explicit, before sending, or reactive. In the latter case, the core first attempts the known path and starts routing when direct delivery fails.
This mode handles path recovery inside the core. A DME may pass through different states:
Queueing and alternative-path selection may change arrival order. Routing provides best-effort reachability according to policies and timeouts; application ordering and deduplication remain properties of the typeset.
Routing and segments
Within a segment, the SID defines the domain in which DINs must be unique. Routing uses this identity to avoid ambiguity between nodes that may share the same underlay.
When two segments merge, both the address space and the graph change. The network must:
- detect any conflicting DINs;
- reconcile the identities of incorporated nodes;
- recalculate their positions in the topology;
- invalidate routes built before the merge;
- propagate the new mappings.
A node's rank, meaning its minimum distance from the root in hop count, provides a stable measure of its logical position. Rebalancing ranks makes it possible to graft one topology onto another while keeping addressing and paths coherent.
Forwarding policies
The local policy decides whether to forward based on the relationship with the sender.
Relevant categories are:
- directly mapped peer;
- peer belonging to the same SID;
- peer known through a route;
- unknown origin.
A closed network accepts mapped nodes. A dynamic mesh enables routed requests and discovery within controlled boundaries. The policy is applied at every hop: rejection by an intermediate node invalidates that path.
Multipath and link selection
A node may know multiple channels to the same peer or multiple next hops to the same destination. Selection may consider:
- driver availability;
- channel state;
- hop count;
- observed latency;
- reliability;
- medium capacity;
- energy policies.
Selection may consider capacity, stability, and hop count. Two hops over stable transports may be preferable to a low-capacity direct link. Routes and drivers may evolve while the DDO remains unchanged.
Routing observability
Network behavior depends on four states:
- locally originated DMEs;
- DMEs received by their destination;
- forwarded DMEs;
- route discoveries started;
- routes expired or invalidated;
- packets delegated, queued, or failed;
- errors by driver and next hop.
These metrics show whether a problem lies in topology knowledge, an intermediate node's policy, or the physical transport.
Transport-independent routing
DaaS-IoT routing keeps communication between DINs stable while links, endpoints, and topology change. The network uses intermediate nodes, crosses different technologies, and rebuilds paths; the application continues to address DINs.
The DDO continues to describe the data; the DME preserves origin and destination; routing selects the hop sequence; drivers perform each physical transfer.