this is a wrapping of weighted round-robin
schedule algorithm, utilizing atomic operation
and cache queue in order to avoid lock latency or the schedule latency. And we have
used an async RwLock
(feature default
or tokio
) to overcome the conflict of select instance and
recalculate queue.
- async interface for tokio
- Atomic operation aimed to provide the best run-time performance
- dynamic insert supported
more detailed documented WrrQueue | Instance
use async_wrr_queue::*;
#[tokio::main]
async fn main() {
let mut queue = WrrQueue::new();
// insert many
queue.insert_many(vec![("a", 1usize), ("b", 2usize)]).await;
// insert one
queue.insert(("c", 3usize)).await;
queue.insert_many(vec![("d", 5usize), ("e", 2usize)]).await;
// schedule!
let mut result = Vec::new();
for _ in 0..30 {
result.push(queue.select().await.unwrap().data().clone());
}
// expected to be this sequence:
assert_eq!(result, Vec::from_iter( [ "d", "c", "b", "d", "e", "d", "c", "a", "d", "b", "e", "c", "d"].into_iter().cycle().take(30)));
}
default
:tokio
tokio
: async interface, usingtokio::sync::RwLock
to guarantee best performanceblocking
: not compatible withtokio
, usingstd::sync::RwLock
for blocking acquire