Capnproto sometimes very slow to send message

Learning
,
1

I am using capnproto to send messages, here is my code:

main.ml

let secret_key = `Ephemeral
	
let start_server id =
	let listen_address = `TCP ("127.0.0.1", 9000 + id) in
	let config = Capnp_rpc_unix.Vat_config.create ~serve_tls:false ~secret_key listen_address in
	let service_id = Capnp_rpc_net.Restorer.Id.public "" in
	let restore = Capnp_rpc_net.Restorer.single service_id (Echo.local id) in
	let+ vat = Capnp_rpc_unix.serve config ~restore in
	let uri = Capnp_rpc_unix.Vat.sturdy_uri vat service_id in
	Fmt.pr "Server running. Connect to URI %S.@." (Uri.to_string uri)

let () =
	Lwt_main.run begin
		let reqs = 1000 in
		let rate = 50. in
		let* () = start_server 0 in
		let* () = start_server 1 in
		let* () = start_server 2 in
    	let conns = Echo.open_conns 3 in
		let* results = Lwt_list.mapi_p
			(fun i msg ->
				let* () = Lwt_unix.sleep ((Float.of_int i) /. rate) in
				Echo.ping_msg (List.nth conns (i mod 3)) msg
			)
			(List.init reqs (fun x -> "ping #" ^ (Int.to_string x)))
		in
		print_results results
	end

echo.ml

let local id =
	let module Echo = Api.Service.Echo in
	Echo.local @@ object
		inherit Echo.service

		method ping_impl params release_param_caps =
			let open Echo.Ping in
			let msg = Params.msg_get params in
      		        Fmt.pr "%d recv: %s@." id msg;
			release_param_caps ();
			Service.return_empty ()
	end


let rec connect service t =
		let* r = Sturdy_ref.connect service in
		match r with
			| Ok conn -> Lwt.return conn
			| Error _ ->
				let* () = Lwt_unix.sleep t in
				Fmt.pr "backed off %f...@." t;
				connect service (t *. 2.) (* binary exponential backoff *)

let open_conn vat id  =
	let uri = Uri.of_string ("capnp://insecure@127.0.0.1:" ^ Int.to_string (id + 9000)) in
	Capnp_rpc_unix.Vat.import_exn vat uri
	
let open_conns nodes =
	let client_vat = Capnp_rpc_unix.client_only_vat () in
	let ids = List.init nodes Fun.id in
	List.map (open_conn client_vat) ids

(* calculate delta between timestamps in seconds s*)
let delta x y =
	let open Base.Int63 in
	(to_float (y - x)) /. 1_000_000_000.

let ping_msg sr msg =
	let ping cap =
		let open Api.Client.Echo.Ping in
		let request, params = Capability.Request.create Params.init_pointer in
		Params.msg_set params msg;
		let t1 = Time_now.nanoseconds_since_unix_epoch () in
		let* () = Capability.call_for_unit_exn cap method_id request in
		let t2 = Time_now.nanoseconds_since_unix_epoch () in
		Lwt.return (delta t1 t2)
	in
	let t1 = Time_now.nanoseconds_since_unix_epoch () in
	let* cap = connect sr 0.01 in
	let t2 = Time_now.nanoseconds_since_unix_epoch () in
	let* t = Capability.with_ref cap ping in
	Lwt.return((t, delta t1 t2))

echo_api.capnp

interface Echo {
	ping @0 (msg :Text) -> ();
}

The time it takes to execute Capability.call_for_unit_exn is normally on the order of 0.1ms, but occasionally it takes on the order of 10ms. What could cause it to execute so slowly sometimes?

2

10ms is a typical time for the garbage collector to do some work.

3

That seems plausible except that I time many other intervals in my program (such as how long it takes to connect) that also take about 0.1ms to execute. The call to send the message is the only one that consistently exhibits this behavior of sometimes taking 10ms so is there anything special about this command that might cause the garbage collector to run?
Also is there anything I can do about it?