Kademlia in Rust

Paper: http://pdos.csail.mit.edu/~petar/papers/maymounkov-kademlia-lncs.pdf

This is a basic (currently WIP) implementation in Rust of the Kademlia distributed hash table, a peer-to-peer information system. It stores both keys and values as character strings, and does not rely on master-slave relations; all nodes in the system are peers.

This is a personal project which I started primarily to learn Rust. Along the way I got to use some deas from my systems programming regarding shared mutable state and programming multi-threaded applications.

Running

RUST_LOG=info cargo run

If started without giving any parameters, a new node with random ID listening on a random local port will be created.

To re-create a given node, you can provide a suitable node ID via the --id parameter. To select the address and/or port a node should bind to, use the --bind parameter.

To connect to another node for bootstrapping off it you must provide both the remote address as well as the remote node's ID via the --bootstrap-addr and --bootstrap-id parameters respectively.

Once a node starts, it will log its information (IP,Port,Key) to stdout PROVIDED THAT RUST_LOG IS SET TO info in the environment.

At this point, you can enter some commands:

put <string key> <string value> ..store value at key
get <string key>                ..lookup value at key
======\/ lower level \/======
p <ip>:<port> <key>  ..pings the node
s <ip>:<port> <key>  ..sends store req to node
fn <ip>:<port> <key> ..sends find_node req to node
fv <ip>:<port> <key> ..sends find_value req to node
ln <key>             ..performs iterative node lookup
lv <string key>      ..performs iterative value lookup

Note that there is a distinction between keys (20-length byte strings) and string keys (arbitrary strings).

Implementation

Each Kademlia node has two large components: the node itself (represented by the Kademlia struct), and the remote procedure call (RPC) facilities, represented by the Rpc struct. They are a bit more tightly coupled than I would like, but I've done my best to keep them separate where possible.

The Kademlia node includes the routing table (K buckets), the store (a basic hash map), and a reference to an Rpc. The Rpc struct allows the node to make RPCs, and also provides a source of incoming requests to the node. The prerequisite to both of these is Rpc::open(), which takes a UdpSocket and a Rust channel Sender, and starts a new thread to parse, mux, and pass along the incoming messages.

Of course, the end user doesn't see all this; they just have to call Kademlia::start() with the appropriate arguments, and they will get back a handle to the node, and this will all happen in the background.

Feedback

I'm certain there are countless mistakes in this project, and I would love to hear how I could improve them (especially with regards to error handling).