mem backend for tx5

Cargo.toml

@@ -50,6 +50,7 @@ sbd-server = "0.0.6-alpha"
 serde = { version = "1.0.160", features = [ "derive", "rc" ] }
 serde_json = { version = "1.0.96", features = [ "preserve_order" ] }
 sha2 = "0.10.6"
+slab = "0.4.9"
 socket2 = { version = "0.5.2", features = [ "all" ] }
 sodoken = "=0.0.901-alpha"
 tempfile = "3.8.0"

crates/tx5/Cargo.toml

@@ -24,6 +24,7 @@ base64 = { workspace = true }
 futures = { workspace = true }
 influxive-otel-atomic-obs = { workspace = true }
 serde = { workspace = true }
+slab = { workspace = true }
 tokio = { workspace = true, features = [ "full" ] }
 tracing = { workspace = true }
 tx5-connection = { workspace = true, default-features = false }

crates/tx5/examples/mem-echo-stress.rs

@@ -0,0 +1,92 @@
+//! Mem backend timings. The hope is to have near ideal zero overhead.
+//! At time of writing, you can see we're pretty close to exactly
+//! proportional to node count for a given machine's resources:
+//!
+//! - 1 nodes, 33941 ops, 0.000028 sec/op, in 1.004478s
+//! - 10 nodes, 174041 ops, 0.000056 sec/op, in 1.016977s
+//! - 100 nodes, 175111 ops, 0.000580 sec/op, in 1.018943s
+//! - 1000 nodes, 160266 ops, 0.006361 sec/op, in 1.019859s
+
+use std::sync::Arc;
+use tx5::{backend::*, *};
+
+#[tokio::main(flavor = "multi_thread")]
+async fn main() {
+    let fake_sig = SigUrl::parse("wss://fake.fake").unwrap();
+
+    let config = Arc::new(Config {
+        backend_module: BackendModule::Mem,
+        ..Default::default()
+    });
+
+    let (listen_ep, mut listen_recv) = Endpoint::new(config.clone());
+    let listen_ep = Arc::new(listen_ep);
+    let listen_url = listen_ep.listen(fake_sig.clone()).await.unwrap();
+
+    tokio::task::spawn(async move {
+        while let Some(evt) = listen_recv.recv().await {
+            match evt {
+                EndpointEvent::ListeningAddressOpen { .. } => (),
+                EndpointEvent::Connected { .. } => (),
+                EndpointEvent::Message { peer_url, .. } => {
+                    let listen_ep = listen_ep.clone();
+                    // Don't hold up our recv loop on the response
+                    tokio::task::spawn(async move {
+                        listen_ep.send(peer_url, Vec::new()).await.unwrap();
+                    });
+                }
+                _ => panic!("unexpected: {evt:?}"),
+            }
+        }
+        panic!("listener task ended");
+    });
+
+    println!("listening at: {listen_url}");
+
+    let (timing_send, mut timing_recv) = tokio::sync::mpsc::unbounded_channel();
+    let mut timing_buf = Vec::new();
+    let mut node_count = 0;
+
+    loop {
+        let start = std::time::Instant::now();
+
+        node_count += 1;
+
+        let config = config.clone();
+        let timing_send = timing_send.clone();
+        let listen_url = listen_url.clone();
+        tokio::task::spawn(async move {
+            let (cli_ep, mut cli_recv) = Endpoint::new(config);
+            let _ = cli_ep.listen(listen_url.to_sig()).await.unwrap();
+
+            loop {
+                let start = std::time::Instant::now();
+                cli_ep.send(listen_url.clone(), Vec::new()).await.unwrap();
+                loop {
+                    let evt = cli_recv.recv().await.unwrap();
+                    match evt {
+                        EndpointEvent::ListeningAddressOpen { .. } => (),
+                        EndpointEvent::Connected { .. } => (),
+                        EndpointEvent::Message { .. } => break,
+                        _ => panic!("unexpected: {evt:?}"),
+                    }
+                }
+                timing_send.send(start.elapsed().as_secs_f64()).unwrap();
+            }
+        });
+
+        tokio::time::sleep(std::time::Duration::from_secs(1)).await;
+
+        timing_recv.recv_many(&mut timing_buf, usize::MAX).await;
+
+        let count = timing_buf.len();
+        let sum: f64 = timing_buf.iter().sum();
+        timing_buf.clear();
+
+        println!(
+            "{node_count} nodes, {count} ops, {:0.6} sec/op, in {:0.6}s",
+            sum / count as f64,
+            start.elapsed().as_secs_f64(),
+        );
+    }
+}

crates/tx5/src/backend.rs

@@ -11,6 +11,8 @@ use tx5_core::deps::serde_json;
 #[cfg(feature = "backend-go-pion")]
 mod go_pion;
 
+mod mem;
+
 /// Backend modules usable by tx5.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
 pub enum BackendModule {
@@ -22,8 +24,8 @@ pub enum BackendModule {
     /// The Webrtc-RS-based backend.
     WebrtcRs,
 
-    /// The mock backend.
-    Mock,
+    /// The mem backend.
+    Mem,
 }
 
 impl Default for BackendModule {
@@ -33,7 +35,7 @@ impl Default for BackendModule {
         return Self::GoPion;
         #[cfg(feature = "backend-webrtc-rs")]
         return Self::WebrtcRs;
-        Self::Mock
+        Self::Mem
     }
 }
 
@@ -45,7 +47,7 @@ impl BackendModule {
             Self::GoPion => go_pion::default_config(),
             #[cfg(feature = "backend-webrtc-rs")]
             Self::WebrtcRs => todo!(),
-            Self::Mock => serde_json::json!({}),
+            Self::Mem => mem::default_config(),
         }
     }
 
@@ -61,7 +63,7 @@ impl BackendModule {
             Self::GoPion => go_pion::connect(config, url, listener).await,
             #[cfg(feature = "backend-webrtc-rs")]
             Self::WebrtcRs => todo!(),
-            Self::Mock => todo!(),
+            Self::Mem => mem::connect(config, url, listener).await,
         }
     }
 }