Crate multiqueue

Expand description

This crate provides a fast mpmc broadcast queue. It’s based on the queue design from the LMAX Disruptor, with a few improvements:

It acts as a futures stream/sink, so you can set up high-performance pipelines
It can dynamically add/remove senders, and each stream can have multiple receivers
It has fast runtime fallbacks for when there’s a single consumer and/or a single producer
It works on 32 bit systems without any performance or capability penalty
In most cases, one can view data written directly into the queue without copying it

In many cases, MultiQueue will be a good replacement for channels and it’s broadcast capabilities can replace more complex concurrency systems with a single queue.

#Queue Model: MultiQueue functions similarly to the LMAX Disruptor from a high level view. There’s an incoming FIFO data stream that is broadcast to a set of subscribers as if there were multiple streams being written to. There are two main differences:

MultiQueue transparently supports switching between single and multiple producers.
Each broadcast stream can be shared among multiple consumers.

The last part makes the model a bit confusing, since there’s a difference between a stream of data and something consuming that stream. To make things worse, each consumer may not actually see each value on the stream. Instead, multiple consumers may act on a single stream each getting unique access to certain elements.

A helpful mental model may be to think about this as if each stream was really just an mpmc queue that was getting pushed to, and the MultiQueue structure just assembled a bunch together behind the scenes. This isn’t the case of course, but it’s helpful for thinking.

An diagram that represents a general use case of the queue where each consumer has unique access to a stream is below - the # stand in for producers and @ stands in for the consumer of each stream, each with a label. The lines are meant to show the data flow through the queue.

. -> #        @-1
.     \      /
.      -> -> -> @-2
.     /      \
. -> #        @-3

This is a pretty standard broadcast queue setup - for each element sent in, it is seen on each stream by that’s streams consumer.

However, in MultiQueue, each logical consumer might actually be demultiplexed across many actual consumers, like below.

. -> #        @-1
.     \      /
.      -> -> -> @-2' (really @+@+@ each compete for a spot)
.     /      \
. -> #        @-3

If this diagram is redrawn with each of the producers sending in a sequenced element (time goes left to right):

. t=1|t=2|    t=3    | t=4|
. 1 -> #              @-1 (1, 2)
.       \            /
.        -> 2 -> 1 -> -> @-2' (really @ (1) + @ (2) + @ (nothing yet))
.       /            \
. 2 -> #              @-3 (1, 2)

If one imagines this as a webserver, the streams for @-1 and @-3 might be doing random webservery work like some logging or metrics gathering and can handle the workload completely on one core, @-2 is doing expensive work handling requests and is split into multiple workers dealing with the data stream.

#MPMC Mode: One might notice that the broadcast queue modes requires that a type be Clone, and the single-reader inplace variants require that a type be Sync as well. This is only required for broadcast queues and not normal mpmc queues, so there’s an mpmc api as well. It doesn’t require that a type be Clone or Sync for any api, and also moves items directly out of the queue instead of cloning them.

#Futures Mode: For both mpmc and broadcast, a futures mode is supported. The datastructures are quite similar to the normal ones, except they implement the Futures Sink/Stream traits for senders and receivers. This comes at a bit of a performance cost, which is why the futures types are separate

#Usage: From the receiving side, this behaves quite similarly to a channel receiver. The .recv function will block until data is available and then return the data.

For senders, there is only .try_send (except for the futures sink, which can park), This is due to performance and api reasons - you should handle backlog instead of just blocking.

§Example: SPSC channel

extern crate multiqueue;

use std::thread;

let (send, recv) = multiqueue::mpmc_queue(10);

let handle = thread::spawn(move || {
    for val in recv {
        println!("Got {}", val);
    }
});

for i in 0..10 {
    send.try_send(i).unwrap();
}

// Drop the sender to close the queue
drop(send);

handle.join();

// prints
// Got 0
// Got 1
// Got 2
// etc

§Example: SPSC broadcasting

extern crate multiqueue;

use std::thread;

let (send, recv) = multiqueue::broadcast_queue(4);
let mut handles = vec![];
for i in 0..2 { // or n
    let cur_recv = recv.add_stream();
    handles.push(thread::spawn(move || {
        for val in cur_recv {
            println!("Stream {} got {}", i, val);
        }
    }));
}

// Take notice that I drop the reader - this removes it from
// the queue, meaning that the readers in the new threads
// won't get starved by the lack of progress from recv
recv.unsubscribe();

for i in 0..10 {
    // Don't do this busy loop in real stuff unless you're really sure
    loop {
        if send.try_send(i).is_ok() {
            break;
        }
    }
}

// Drop the sender to close the queue
drop(send);

for t in handles {
    t.join();
}

// prints along the lines of
// Stream 0 got 0
// Stream 0 got 1
// Stream 1 got 0
// Stream 0 got 2
// Stream 1 got 1
// etc

§Example: SPMC broadcast

extern crate multiqueue;

use std::thread;

let (send, recv) = multiqueue::broadcast_queue(4);

let mut handles = vec![];

for i in 0..2 { // or n
    let cur_recv = recv.add_stream();
    for j in 0..2 {
        let stream_consumer = cur_recv.clone();
        handles.push(thread::spawn(move || {
            for val in stream_consumer {
                println!("Stream {} consumer {} got {}", i, j, val);
            }
        }));
    }
    // cur_recv is dropped here
}

// Take notice that I drop the reader - this removes it from
// the queue, meaning that the readers in the new threads
// won't get starved by the lack of progress from recv
recv.unsubscribe();

for i in 0..10 {
    // Don't do this busy loop in real stuff unless you're really sure
    loop {
        if send.try_send(i).is_ok() {
            break;
        }
    }
}
drop(send);

for t in handles {
    t.join();
}

// prints along the lines of
// Stream 0 consumer 1 got 2
// Stream 0 consumer 0 got 0
// Stream 1 consumer 0 got 0
// Stream 0 consumer 1 got 1
// Stream 1 consumer 1 got 1
// Stream 1 consumer 0 got 2
// etc

// some join mechanics here

§Example: Usage menagerie

extern crate multiqueue;

use std::thread;

let (send, recv) = multiqueue::broadcast_queue(4);
let mut handles = vec![];

// start like before
for i in 0..2 { // or n
    let cur_recv = recv.add_stream();
    for j in 0..2 {
        let stream_consumer = cur_recv.clone();
        handles.push(thread::spawn(move ||
            for val in stream_consumer {
                println!("Stream {} consumer {} got {}", i, j, val);
            }
        ));
    }
    // cur_recv is dropped here
}

// On this stream, since there's only one consumer,
// the receiver can be made into a UniReceiver
// which can view items inline in the queue
let single_recv = recv.add_stream().into_single().unwrap();

handles.push(thread::spawn(move ||
    for val in single_recv.iter_with(|item_ref| 10 * *item_ref) {
        println!("{}", val);
    }
));

// Same as above, except this time we just want to iterate until the receiver is empty
let single_recv_2 = recv.add_stream().into_single().unwrap();

handles.push(thread::spawn(move ||
    for val in single_recv_2.try_iter_with(|item_ref| 10 * *item_ref) {
        println!("{}", val);
    }
));

// Take notice that I drop the reader - this removes it from
// the queue, meaning that the readers in the new threads
// won't get starved by the lack of progress from recv
recv.unsubscribe();

// Many senders to give all the receivers something
for _ in 0..3 {
    let cur_send = send.clone();
    handles.push(thread::spawn(move ||
        for i in 0..10 {
            loop {
                if cur_send.try_send(i).is_ok() {
                    break;
                }
            }
        }
    ));
}
drop(send);

for t in handles {
   t.join();
}

Modules§

wait: This module contains the waiting strategies used by the queue when there is no data left. Users should not find themselves directly accessing these except for construction unless a custom Wait is being written.

Structs§

BroadcastFutReceiver: This is the futures-compatible version of BroadcastReceiver It implements Stream
BroadcastFutSender: This is the futures-compatible version of BroadcastSender It implements Sink
BroadcastFutUniReceiver: This is the futures-compatible version of BroadcastUniReceiver It implements Stream and behaves like the iterator would. To use a different function must transform itself into a different BroadcastFutUniRecveiver use transform_operation
BroadcastReceiver: This class is the receiving half of the broadcast MultiQueue. Within each stream, it supports both single and multi consumer modes with competitive performance in each case. It supports blocking and nonblocking read modes as well as being the conduit for adding new streams.
BroadcastSender: This class is the sending half of the broadcasting MultiQueue. It supports both single and multi consumer modes with competitive performance in each case. It only supports nonblocking writes (the futures sender being an exception) as well as being the conduit for adding new writers.
BroadcastUniReceiver: This class is similar to the receiver, except it ensures that there is only one consumer for the stream it owns. This means that one can safely view the data in-place with the recv_view method family and avoid the cost of copying it. If there’s only one receiver on a stream, it can be converted into a BroadcastUniInnerRecv
MPMCFutReceiver: This is the futures-compatible version of MPMCReceiver It implements Stream
MPMCFutSender: This is the futures-compatible version of MPMCSender It implements Sink
MPMCFutUniReceiver: This is the futures-compatible version of MPMCUniReceiver It implements Stream and behaves like the iterator would. To use a different function must transform itself into a different UniRecveiver use transform_operation
MPMCReceiver: This is the receiving end of a standard mpmc view of the queue It functions similarly to the BroadcastReceiver execpt there is only ever one stream. As a result, the type doesn’t need to be clone
MPMCSender: This class is the sending half of the mpmc MultiQueue. It supports both single and multi consumer modes with competitive performance in each case. It only supports nonblocking writes (the futures sender being an exception) as well as being the conduit for adding new writers.
MPMCUniReceiver: This is the receiving end of a standard mpmc view of the queue for when it’s statically know that there is only one receiver. It functions similarly to the BroadcastUniReceiver execpt there is only ever one stream. As a result, the type doesn’t need to be clone or sync

Functions§

broadcast_fut_queue: Futures variant of broadcast_queue - datastructures implement Sink + Stream at a minor (~30 ns) performance cost to BlockingWait
broadcast_queue: Creates a (BroadcastSender, BroadcastReceiver) pair with a capacity that’s the next power of two >= the given capacity
broadcast_queue_with: Creates a (BroadcastSender, BroadcastReceiver) pair with a capacity that’s the next power of two >= the given capacity and the specified wait strategy
mpmc_fut_queue: Futures variant of mpmc_queue - datastructures implement Sink + Stream at a minor (~30 ns) performance cost to BlockingWait
mpmc_queue: Creates a (MPMCSender, MPMCReceiver) pair with a capacity that’s the next power of two >= the given capacity
mpmc_queue_with

Crate multiqueueCopy item path