The grid operator issues a command. The meter executes it now.

Legacy smart meters operate on a poll cycle. The head-end asks, the meter answers — on schedule. Commands wait in queue. Disconnect orders arrive hours later, sometimes days. The grid cannot respond to what it cannot reach in real time.

DLMS/COSEM is a request-response protocol. The utility polls. The meter answers. Between polls, the meter is deaf. A tamper event, a load spike, a payment — none of these reach the operator until the next scheduled read. The meter has the data. The operator does not.

MQTTS changes the topology

MQTTS is publish-subscribe. The meter maintains a persistent, lightweight, encrypted connection. Events publish the moment they occur. Commands arrive the moment they are issued. There is no poll cycle. There is no queue.

Disconnect, reconnect, load-limit, firmware update, tariff change — every command executes in real time over the existing connection. The grid operator has continuous, bidirectional control of every device on the network.

Edge compute, not cloud dependency

Critical operations — tamper detection, credit enforcement, load management — execute locally on the meter. The cloud is for analytics and fleet management. When the network drops, the meter continues to operate, detect, and enforce. When the network returns, the event backlog publishes automatically.

This is not an incremental improvement over poll-based architecture. It is a different paradigm. The grid stops being a collection of endpoints that report periodically and becomes a distributed, continuously responsive system.