Response Timing

Every /rpc response includes a timing object that breaks down where time was spent. Use it to diagnose latency and understand the full round-trip cost of a request.

TIP

All timings are in milliseconds (ms). 1000 ms = 1 second, so a 200 ms response is 0.2 s.

{
  "ok": true,
  "requestId": null,
  "result": "hello",
  "error": null,
  "timing": {
    "decryptMs": 1,
    "connectMs": 0,
    "execMs": 9,
    "proxyMs": 10
  }
}

Fields

Field	Measured at	What it captures
decryptMs	Proxy node	Time to decrypt the encrypted service credentials that the gateway sent. Includes the AES-GCM operation. Normally < 4 ms.
connectMs	Proxy node	Time to acquire a connection from the per-service connection pool. 0 ms means a pooled connection was reused immediately. A non-zero value means the pool was empty and a new TCP connection had to be opened to your database.
execMs	Proxy node	Time to run the actual command against your database and receive the result. This is pure database + network time between the proxy node and your database host.
proxyMs	Proxy node	Total wall-clock time spent inside the proxy node for this request. Equals roughly decryptMs + connectMs + execMs plus a small internal overhead (JSON encoding, etc).

How to read total latency

A request travels through two hops:

Client → [gateway] → [proxy node] → [your database]
          └─ gateway stages ─┘   └─ node timing ──┘

• proxyMs — the node's share of the round-trip (decrypt + connect + exec).
• Gateway overhead — everything else: API key auth, signature verification, request routing, and JWT signing before the request reaches the node.
• Network — transit between client→gateway and gateway→node is not directly measured but is totalMs − gatewayOverhead − proxyMs.

Optimising

connectMs is high	Your connection pool is cold or undersized. Reuse the zedgi client across requests (module-level singleton) so the pool stays warm.
execMs is high	The query itself is slow. Add an index or optimise the SQL.
decryptMs > 5 ms	Rare; indicates CPU pressure on the proxy node.
Total > 500 ms but proxyMs < 50 ms	Latency is in the gateway or network rather than your database.