For a typical golang network program, you would first conn := lis.Accept() to get a connection and go func(net.Conn) to start a goroutine for handling the incoming data, then you would buf:=make([]byte, 4096) to allocate some buffer and finally waits on conn.Read(buf).
For a server holding >10K connections with frequent short messages(e.g. < 512B), cost for context switching is much more expensive than receiving message(a context switch needs at least 1000 CPU cycles or 600ns on 2.1GHz).
And by eliminating one goroutine per one connection scheme with Edge-Triggered IO Multiplexing, the 2KB(R)+2KB(W) per connection goroutine stack can be saved. By using internal swap buffer, buf:=make([]byte, 4096) can be saved(at the cost of performance).
gaio is an proactor pattern networking library satisfy both memory constraints and performance goals.
- Tested in High Frequency Trading for handling HTTP requests for 30K~40K RPS on a single HVM server.
- Designed for >C10K concurrent connections, maximized parallelism, and nice single connection throughput.
- Read(ctx, conn, buffer) can be called with
nilbuffer to make use of internal swap buffer. - Non-intrusive design, this library works with net.Listener and net.Conn. (with syscall.RawConn support), easy to be integrated into your existing software.
- Amortized context switching cost for tiny messages, able to handle frequent chat message exchanging.
- Application can decide when to delegate net.Conn to
gaio, for example, you can delegate net.Conn togaioafter some handshaking procedure, or having some net.TCPConn settings done. - Application can decide when to submit read or write requests, per-connection back-pressure can be propagated to peer to slow down sending. This features is particular useful to transmit data from A to B via gaio, which B is slower than A.
- Tiny, around 1000 LOC, easy to debug.
- Support for Linux, BSD.
- Once you submit an async read/write requests with related net.Conn to gaio.Watcher, this conn will be delegated to
gaio.Watcherat first submit. Future use of this conn like conn.Read or conn.Write will return error, but TCP properties set bySetReadBuffer(),SetWriteBuffer(),SetLinger(),SetKeepAlive(),SetNoDelay()will be inherited. - If you decide not to use this connection anymore, you could call Watcher.Free(net.Conn) to close socket and free related resources immediately.
- If you forget to call Watcher.Free(net.Conn), runtime garbage collector will cleanup related system resources if nowhere in the system holds the net.Conn.
- If you forget to call Watcher.Close(), runtime garbage collector will cleanup ALL related system resources if nowhere in the system holds this
Watcher. - For connection Load-Balance, you can create multiple gaio.Watcher with your own strategy to distribute net.Conn.
- For acceptor Load-Balance, you can use go-reuseport as the listener.
- For read requests submitted with 'nil' buffer, the returning
[]bytefromWatcher.WaitIO()is SAFE to use before next call to Watcher.WaitIO() returned.
package main
import (
"log"
"net"
"github.com/xtaci/gaio"
)
// this goroutine will wait for all io events, and sents back everything it received
// in async way
func echoServer(w *gaio.Watcher) {
for {
// loop wait for any IO events
results, err := w.WaitIO()
if err != nil {
log.Println(err)
return
}
for _, res := range results {
switch res.Operation {
case gaio.OpRead: // read completion event
if res.Error == nil {
// send back everything, we won't start to read again until write completes.
// submit an async write request
w.Write(nil, res.Conn, res.Buffer[:res.Size])
}
case gaio.OpWrite: // write completion event
if res.Error == nil {
// since write has completed, let's start read on this conn again
w.Read(nil, res.Conn, res.Buffer[:cap(res.Buffer)])
}
}
}
}
}
func main() {
w, err := gaio.NewWatcher()
if err != nil {
log.Fatal(err)
}
defer w.Close()
go echoServer(w)
ln, err := net.Listen("tcp", "localhost:0")
if err != nil {
log.Fatal(err)
}
log.Println("echo server listening on", ln.Addr())
for {
conn, err := ln.Accept()
if err != nil {
log.Println(err)
return
}
log.Println("new client", conn.RemoteAddr())
// submit the first async read IO request
err = w.Read(nil, conn, make([]byte, 128))
if err != nil {
log.Println(err)
return
}
}
}Push server
package mainpackage main
import (
"fmt"
"log"
"net"
"time"
"github.com/xtaci/gaio"
)
func main() {
// by simply replace net.Listen with reuseport.Listen, everything is the same as in push-server
// ln, err := reuseport.Listen("tcp", "localhost:0")
ln, err := net.Listen("tcp", "localhost:0")
if err != nil {
log.Fatal(err)
}
log.Println("pushing server listening on", ln.Addr(), ", use telnet to receive push")
// create a watcher
w, err := gaio.NewWatcher()
if err != nil {
log.Fatal(err)
}
// channel
ticker := time.NewTicker(time.Second)
chConn := make(chan net.Conn)
chIO := make(chan gaio.OpResult)
// watcher.WaitIO goroutine
go func() {
for {
results, err := w.WaitIO()
if err != nil {
log.Println(err)
return
}
for _, res := range results {
chIO <- res
}
}
}()
// main logic loop, like your program core loop.
go func() {
var conns []net.Conn
for {
select {
case res := <-chIO: // receive IO events from watcher
if res.Error != nil {
continue
}
conns = append(conns, res.Conn)
case t := <-ticker.C: // receive ticker events
push := []byte(fmt.Sprintf("%s\n", t))
// all conn will receive the same 'push' content
for _, conn := range conns {
w.Write(nil, conn, push)
}
conns = nil
case conn := <-chConn: // receive new connection events
conns = append(conns, conn)
}
}
}()
// this loop keeps on accepting connections and send to main loop
for {
conn, err := ln.Accept()
if err != nil {
log.Println(err)
return
}
chConn <- conn
}
}For complete documentation, see the associated Godoc.
| Test Case | Throughput test with 64KB buffer |
|---|---|
| Description | A client keep on sending 64KB bytes to server, server keeps on reading and sending back whatever it received, the client keeps on receiving whatever the server sent back until all bytes received successfully |
| Command | go test -v -run=^$ -bench Echo |
| Macbook Pro | 1695.27 MB/s 518 B/op 4 allocs/op |
| Linux AMD64 | 1883.23 MB/s 518 B/op 4 allocs/op |
| Raspberry Pi4 | 354.59 MB/s 334 B/op 4 allocs/op |
| Test Case | 8K concurrent connection echo test |
|---|---|
| Description | Start 8192 clients, each client send 1KB data to server, server keeps on reading and sending back whatever it received, the client keeps on receiving whatever the server sent back until all bytes received successfully. |
| Command | go test -v -run=8k |
| Macbook Pro | 1.09s |
| Linux AMD64 | 0.94s |
| Raspberry Pi4 | 2.09s |
X -> number of concurrent connections, Y -> time of completion in seconds
Best-fit values
Slope 8.613e-005 ± 5.272e-006
Y-intercept 0.08278 ± 0.03998
X-intercept -961.1
1/Slope 11610
95% Confidence Intervals
Slope 7.150e-005 to 0.0001008
Y-intercept -0.02820 to 0.1938
X-intercept -2642 to 287.1
Goodness of Fit
R square 0.9852
Sy.x 0.05421
Is slope significantly non-zero?
F 266.9
DFn,DFd 1,4
P Value < 0.0001
Deviation from horizontal? Significant
Data
Number of XY pairs 6
Equation Y = 8.613e-005*X + 0.08278
gaio source code is available under the MIT License.
- https://zhuanlan.zhihu.com/p/102890337 -- gaio小记
- golang/go#15735 -- net: add mechanism to wait for readability on a TCPConn
- https://en.wikipedia.org/wiki/C10k_problem -- C10K
- https://golang.org/src/runtime/netpoll_epoll.go -- epoll in golang
- https://www.freebsd.org/cgi/man.cgi?query=kqueue&sektion=2 -- kqueue
- https://idea.popcount.org/2017-02-20-epoll-is-fundamentally-broken-12/ -- epoll is fundamentally broken
- https://en.wikipedia.org/wiki/Transmission_Control_Protocol#Flow_control -- TCP Flow Control
- http://www.idc-online.com/technical_references/pdfs/data_communications/Congestion_Control.pdf -- Back-pressure
Stable