Struct kitsune_p2p_dht::arq::PeerViewQ

pub struct PeerViewQ {
    pub topo: Topology,
    pub skip_index: Option<usize>,
    /* private fields */
}

The Quantized PeerView

Fields

topo: Topology

The topology of the network space

skip_index: Option<usize>

Omit the arq at this index from all peer considerations. Useful for tests which update all arqs, without needing to construct a new PeerView for each arq needing to be updated

Implementations

impl PeerViewQ

pub fn new(topo: Topology, strat: ArqStrat, peers: Vec<Arq>) -> Self

Constructor

pub fn actual_coverage(&self) -> f64

The actual coverage of all arcs in this view. TODO: this only makes sense when the view contains all agents in the DHT. So, it’s more useful for testing. Probably want to tease out some concept of a test DHT from this.

pub fn extrapolated_coverage(&self, filter: &Arq) -> f64

Extrapolate the coverage of the entire network from our local view.

pub fn extrapolated_coverage_and_filtered_count( &self, filter: &Arq ) -> (f64, usize)

Return the extrapolated coverage and the number of arqs which match the filter. These two are complected together simply for efficiency’s sake, to minimize computation

TODO: this probably will be rewritten when PeerView is rewritten to have the filter baked in.

pub fn raw_coverage(&self, filter: &Arq) -> f64

Compute the total coverage observed within the filter interval.

pub fn update_arq(&self, arq: &mut Arq) -> bool

Mutate the arq to its ideal target

pub fn slack_factor(&self, cov: f64, num_peers: usize) -> f64

The “slacker” factor. If our observed coverage is significantly greater than the number of peers we see, it’s an indication that we may need to pick up more slack.

This check helps balance out stable but unequitable situations where all peers have a similar estimated coverage, but some peers are holding much more than others.

pub fn update_arq_with_stats(&self, arq: &mut Arq) -> UpdateArqStats

Take an arq and potentially resize and requantize it based on this view.

This represents an iterative step towards the ideal coverage, based on the observed coverage. This makes many assumptions, including:

• this arc resizing algorithm is a good one, namely that the coverage at any point of the DHT is close to the target range
• all other peers are following the same algorithm
• if we see a change that we need to make, we assume that a number of peers are about to make a similar change, and that number is on average the same as our target coverage

More detail on these assumptions here: https://hackmd.io/@hololtd/r1IAIbr5Y/https%3A%2F%2Fhackmd.io%2FK_fkBj6XQO2rCUZRRL9n2g TODO: make the above link to something publicly available, preferably in the repo

pub fn power_stats(&self, topo: &Topology, filter: &Arq) -> PowerStats

Gather statistics about the power levels of all arqs in this view. Used to make inferences about what your neighbors are up to.

Trait Implementations

impl From<PeerViewQ> for PeerView

fn from(original: PeerViewQ) -> PeerView

Converts to this type from the input type.