net.tomp2p.replication

Class AutoReplication

java.lang.Object

net.tomp2p.replication.AutoReplication

All Implemented Interfaces:

PeerMapChangeListener, ReplicationFactor

public class AutoReplication
extends Object
implements PeerMapChangeListener, ReplicationFactor

AutomaticReplication class observes the local network conditions and predicts the next departure count based on previous and current peer churn behaviors. This predicted value and given reliability are used to calculate the replication factor

Author:

Maxat Pernebayev, Thomas Bocek

Constructor Summary

Constructors

Constructor and Description
AutoReplication(Peer peer) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
static double	bestSmoothingFactor(ArrayList<Integer> x, ArrayList<Double> y) The interval length which results in maximum value of R squared is considered to be the best solution for the smoothing factor.
static double	choose(int n, int k) nCR (ohne Zurücklegen, ohne Beachtung der Reihenfolge) http://stackoverflow.com/questions/1678690/what-is-a-good-way-to- implement-choose-notation-in-java
static double	ema(ArrayList<Integer> observations, ArrayList<Double> emas) Exponential Moving Average with Dynamic Smoothing Factor
static double	linearRegression(ArrayList<Integer> x, ArrayList<Double> y, int n) LinearRegression, adapted from: http://introcs.cs.princeton.edu/java/97data/LinearRegression.java.html
int	maxReplicationFactor()
AutoReplication	maxReplicationFactor(int maxReplicationFactor)
int	minReplicationFactor()
AutoReplication	minReplicationFactor(int minReplicationFactor)
int	observationLength()
AutoReplication	observationLength(int observationLength)
void	peerInserted(PeerAddress peerAddress, boolean verified)
int	peerMapSize()
void	peerRemoved(PeerAddress peerAddress, PeerStatatistic storedPeerAddress)
void	peerUpdated(PeerAddress peerAddress, PeerStatatistic storedPeerAddress)
static int	predictedValue(ArrayList<Integer> observations, double currentAverage) Exponential Moving Average with Dynamic Smoothing Factor plus Deviation
double	reliability()
AutoReplication	reliability(double reliability)
int	replicationFactor()
static int	replicationFactor(int m, int n, double r, int minReplicationFactor, int maxReplicationFactor) Calculate the replication factor given the reliability within a churn interval and the (predicted) number of peers that left the system.
static int	replicationFactor2(int m, int n, double r, int minReplicationFactor, int maxReplicationFactor) Calculate the replication factor given the reliability within a churn interval and the (predicted) number of peers that left the system.
AutoReplication	shutdown()
static double	standardDeviation(ArrayList<Integer> range, double average) Calculation of sample standard deviation for a set of values.
AutoReplication	start()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AutoReplication

public AutoReplication(Peer peer)

Constructor.

Parameters:

reliability - The reliability

peerMap - The map of my neighbors

Method Detail

start

public AutoReplication start()

shutdown

public AutoReplication shutdown()

peerInserted

public void peerInserted(PeerAddress peerAddress,
                boolean verified)

Specified by:

peerInserted in interface PeerMapChangeListener

peerUpdated

public void peerUpdated(PeerAddress peerAddress,
               PeerStatatistic storedPeerAddress)

Specified by:

peerUpdated in interface PeerMapChangeListener

peerRemoved

public void peerRemoved(PeerAddress peerAddress,
               PeerStatatistic storedPeerAddress)

Specified by:

peerRemoved in interface PeerMapChangeListener

reliability

public AutoReplication reliability(double reliability)

reliability

public double reliability()

minReplicationFactor

public AutoReplication minReplicationFactor(int minReplicationFactor)

minReplicationFactor

public int minReplicationFactor()

maxReplicationFactor

public AutoReplication maxReplicationFactor(int maxReplicationFactor)

maxReplicationFactor

public int maxReplicationFactor()

observationLength

public AutoReplication observationLength(int observationLength)

observationLength

public int observationLength()

peerMapSize

public int peerMapSize()

Returns:

The number of neighbor peers.

replicationFactor

public int replicationFactor()

Specified by:

replicationFactor in interface ReplicationFactor

Returns:

The replication factor.

replicationFactor

public static int replicationFactor(int m,
                    int n,
                    double r,
                    int minReplicationFactor,
                    int maxReplicationFactor)

Calculate the replication factor given the reliability within a churn interval and the (predicted) number of peers that left the system.

Parameters:

m - (predicted) number of peers that left the system

n - number of all peers

r - reliability

minReplicationFactor - the minimal replication factor, not considering the reliability

maxReplicationFactor - the maximal replication factor, not considering the reliability

Returns:

the replication factor for the desired reliability rate

choose

public static double choose(int n,
            int k)

nCR (ohne Zurücklegen, ohne Beachtung der Reihenfolge) http://stackoverflow.com/questions/1678690/what-is-a-good-way-to- implement-choose-notation-in-java

Parameters:

n -

k -

Returns:

replicationFactor2

public static int replicationFactor2(int m,
                     int n,
                     double r,
                     int minReplicationFactor,
                     int maxReplicationFactor)

Calculate the replication factor given the reliability within a churn interval and the (predicted) number of peers that left the system. Faster than replicationFactor as we only use one for loop.

Parameters:

m - (predicted) number of peers that left the system

n - number of all peers

r - reliability

minReplicationFactor - the minimal replication factor, not considering the reliability

maxReplicationFactor - the maximal replication factor, not considering the reliability

Returns:

the replication factor for the desired reliability rate

bestSmoothingFactor

public static double bestSmoothingFactor(ArrayList<Integer> x,
                         ArrayList<Double> y)

The interval length which results in maximum value of R squared is considered to be the best solution for the smoothing factor.

Parameters:

x - The array of independent values, i.e. observations

y - The array of dependent values, i.e. averages

Returns:

Smoothing factor alpha, which is based on the interval length which results in maximum value of R squared.

linearRegression

public static double linearRegression(ArrayList<Integer> x,
                      ArrayList<Double> y,
                      int n)

LinearRegression, adapted from: http://introcs.cs.princeton.edu/java/97data/LinearRegression.java.html

Parameters:

x - The array of independent values, i.e. observations

y - The array of dependent values, i.e. averages

interval - How much values should be considered

Returns:

R squared - defines how much percent of dependent values can be explained by independent values. The closer R squared is to 1, the better relationship between dependent and independent values.

ema

public static double ema(ArrayList<Integer> observations,
         ArrayList<Double> emas)

Exponential Moving Average with Dynamic Smoothing Factor

Parameters:

observations - Array that contains peer departures numbers.

emas - Array that contains values of exponential moving averages

interval - The length of interval that is used to calculate smoothing factor

Returns:

The exponential moving average for given observations of peer departures.

standardDeviation

public static double standardDeviation(ArrayList<Integer> range,
                       double average)

Calculation of sample standard deviation for a set of values. Inspired by: http://introcs.cs.princeton.edu/java/stdlib/StdStats.java.html

Parameters:

range - A set of values

currentAverage - Average value for the set

Returns:

The standard deviation

predictedValue

public static int predictedValue(ArrayList<Integer> observations,
                 double currentAverage)

Exponential Moving Average with Dynamic Smoothing Factor plus Deviation

Parameters:

observations - Array that contains recent peer departures numbers.

currentAverage - Current exponential moving average

interval - The length of interval that is used to calculate deviation

Returns:

The exponential moving average plus deviation for a given observation of peer departures. This value is regarded as a prediction value for the next second.