node-replication

The node-replication library is used to implement concurrent and replicated versions of (single-threaded) data structures by combining a set of different concurrency techniques: flat combining, operation logging, readers-writer locks, and partitioning.

Implementations

The crate contains two variants of node-replication: NR for transforming sequential data-structures and CNR for already concurrent or partitionable data structures.

nr converts a sequental data structure to its NUMA-aware concurrent version.
cnr converts a concurrent (or partitioned) data structure to its NUMA-aware concurrent version.

Background

The implementation of NR and a detailed evaluation can be found in this paper: Black-box Concurrent Data Structures for NUMA Architectures which explains the NR library in more details.
The implementation of CNR and a detailed evaluation can be found in this paper: NrOS: Effective Replication and Sharing in an Operating System. The NrOS documentation also has a brief overview of NR and CNR as these are the main concurrency mechanisms in the nrkernel.
The NR code in this repo was formally verified by porting it to Dafny, including a proof that it ensures linearizability.

How does it work (NR)

To replicate a single-threaded data structure, one needs to implement Dispatch (from node-replication). As an example, we implement Dispatch for the single-threaded HashMap from std.

#![feature(generic_associated_types)]
use std::collections::HashMap;
use node_replication::nr::Dispatch;

/// The node-replicated hashmap uses a std hashmap internally.
#[derive(Default)]
struct NrHashMap {
    storage: HashMap<u64, u64>,
}

/// We support mutable put operation on the hashmap.
#[derive(Clone, Debug, PartialEq)]
enum Modify {
    Put(u64, u64),
}

/// We support an immutable read operation to lookup a key from the hashmap.
#[derive(Clone, Debug, PartialEq)]
enum Access {
    Get(u64),
}

/// The Dispatch traits executes `ReadOperation` (our Access enum)
/// and `WriteOperation` (our `Modify` enum) against the replicated
/// data-structure.
impl Dispatch for NrHashMap {
    type ReadOperation<'rop> = Access;
    type WriteOperation = Modify;
    type Response = Option<u64>;

    /// The `dispatch` function applies the immutable operations.
    fn dispatch<'rop>(&self, op: Self::ReadOperation<'rop>) -> Self::Response {
        match op {
            Access::Get(key) => self.storage.get(&key).map(|v| *v),
        }
    }

    /// The `dispatch_mut` function applies the mutable operations.
    fn dispatch_mut(&mut self, op: Self::WriteOperation) -> Self::Response {
        match op {
            Modify::Put(key, value) => self.storage.insert(key, value),
        }
    }
}

The full example (using HashMap as the underlying data-structure) can be found here. To run, execute: cargo run --example nr_hashmap

How does it perform (NR)

The library often makes your single-threaded implementation work better than, or competitive with fine-grained locking or lock free implementations of the same data-structure.

It works especially well if

Your data-structure exceeds the L3 cache-size of your system (you may not see any gain from replication if your data can always remain in the cache).
All your threads need to issue mixed, mutable and immutable operations (if not alternative techniques like RCU may work better).
You have enough DRAM to take advantage of the replication (i.e., it's typically best to use one replica per-NUMA node which means your original memory foot-print is multiplied with the amount of NUMA nodes in the system).

As an example, the following benchmark uses Rust' the hash-table with the Dispatch implementation from above (nr), and compares it against concurrent hash table implementations from crates.io (chashmap, dashmap, flurry), a HashMap protected by an RwLock (std), and urcu.

The figures show a benchmark using hash tables pre-filled with 67M entires (8 byte keys and values) and uses a uniform key distribution for operations. On the left graphs, different write ratios (0%, 10% and 80%) are shown. On the right graph, we vary the write ratio (x-axis) with 192 threads. The system has 4 NUMA nodes, so it uses 4 replicas (at x=96, a replica gets added every 24 cores). After x=96, the remaining hyper-threads are used.

Compile the library

The library works with no_std and currently requires a nightly rust compiler.

cargo build

Add it as a dependency in your Cargo.toml:

node-replication = "*"

The code should currently be treated as an early release and is still work in progress.

Testing

There are a range of unit tests as part of the implementation and a few integration tests that check various aspects of the implementations.

You can run the tests by executing: cargo test

Benchmarks

The benchmarks (and how to execute them) are explained in more detail in the benches folder.

Supported Platforms

The code should be treated as an early release and is still work in progress. In its current form, the library is only known to work on x86-64 platforms (other platforms will require some changes and are untested).

Contributing

The node-replication project team welcomes contributions from the community. If you wish to contribute code and you have not signed our contributor license agreement (CLA), our bot will update the issue when you open a Pull Request. For any questions about the CLA process, please refer to our FAQ.

vmware / node-replication

readme