org.apache.cassandra.utils

Class MerkleTree

java.lang.Object

org.apache.cassandra.utils.MerkleTree

public class MerkleTree
extends java.lang.Object

A MerkleTree implemented as a binary tree. A MerkleTree is a full binary tree that represents a perfect binary tree of depth 'hashdepth'. In a perfect binary tree, each leaf contains a sequentially hashed range, and each inner node contains the binary hash of its two children. In the MerkleTree, many ranges will not be split to the full depth of the perfect binary tree: the leaves of this tree are Leaf objects, which contain the computed values of the nodes that would be below them if the tree were perfect. The hash values of the inner nodes of the MerkleTree are calculated lazily based on their children when the hash of a range is requested with hash(range). Inputs passed to TreeRange.validate should be calculated using a very secure hash, because all hashing internal to the tree is accomplished using XOR. If two MerkleTrees have the same hashdepth, they represent a perfect tree of the same depth, and can always be compared, regardless of size or splits.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	MerkleTree.RowHash Hash value representing a row, to be used to pass hashes to the MerkleTree.
static class	MerkleTree.TreeRange The public interface to a range in the tree.
static class	MerkleTree.TreeRangeIterator Returns the leaf (range) of a given tree in increasing order.

Field Summary

Fields

Modifier and Type	Field and Description
static byte	RECOMMENDED_DEPTH

Constructor Summary

Constructors

Constructor and Description
MerkleTree(IPartitioner partitioner, Range<Token> range, int hashdepth, long maxsize)

Method Summary

Modifier and Type	Method and Description
static MerkleTree	deserialize(DataInputPlus in, boolean offHeapRequested, int version)
static MerkleTree	deserialize(DataInputPlus in, int version)
static java.util.List<MerkleTree.TreeRange>	difference(MerkleTree ltree, MerkleTree rtree)
boolean	equals(java.lang.Object other)
static int	estimatedMaxDepthForBytes(IPartitioner partitioner, long numBytes, int bytesPerHash) Estimate the allowable depth while keeping the resulting heap usage of this tree under the provided number of bytes.
MerkleTree.TreeRange	get(Token t) For testing purposes.
void	init() Initializes this tree by splitting it until hashdepth is reached, or until an additional level of splits would violate maxsize.
long	maxsize()
void	maxsize(long maxsize)
IPartitioner	partitioner()
void	release()
long	rowCount()
void	serialize(DataOutputPlus out, int version)
long	serializedSize(int version)
long	size() The number of distinct ranges contained in this tree.
boolean	split(Token t) Splits the range containing the given token, if no tree limits would be violated.
java.lang.String	toString()

Methods inherited from class java.lang.Object

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

Field Detail

RECOMMENDED_DEPTH

public static final byte RECOMMENDED_DEPTH

See Also:

Constant Field Values

Constructor Detail

MerkleTree

public MerkleTree(IPartitioner partitioner,
                  Range<Token> range,
                  int hashdepth,
                  long maxsize)

Parameters:

partitioner - The partitioner in use.

range - the range this tree covers

hashdepth - The maximum depth of the tree. 100/(2^depth) is the % of the key space covered by each subrange of a fully populated tree.

maxsize - The maximum number of subranges in the tree.

Method Detail

init

public void init()

Initializes this tree by splitting it until hashdepth is reached, or until an additional level of splits would violate maxsize. NB: Replaces all nodes in the tree, and always builds on the heap

release

public void release()

partitioner

public IPartitioner partitioner()

size

public long size()

The number of distinct ranges contained in this tree. This is a reasonable measure of the memory usage of the tree (assuming 'this.order' is significant).

maxsize

public long maxsize()

maxsize

public void maxsize(long maxsize)

difference

public static java.util.List<MerkleTree.TreeRange> difference(MerkleTree ltree,
                                                              MerkleTree rtree)

Parameters:

ltree - First tree.

rtree - Second tree.

Returns:

A list of the largest contiguous ranges where the given trees disagree.

split

public boolean split(Token t)

Splits the range containing the given token, if no tree limits would be violated. If the range would be split to a depth below hashdepth, or if the tree already contains maxsize subranges, this operation will fail.

Returns:

True if the range was successfully split.

rowCount

public long rowCount()

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

equals

public boolean equals(java.lang.Object other)

Overrides:

equals in class java.lang.Object

serialize

public void serialize(DataOutputPlus out,
                      int version)
               throws java.io.IOException

Throws:

java.io.IOException

serializedSize

public long serializedSize(int version)

deserialize

public static MerkleTree deserialize(DataInputPlus in,
                                     int version)
                              throws java.io.IOException

Throws:

java.io.IOException

deserialize

public static MerkleTree deserialize(DataInputPlus in,
                                     boolean offHeapRequested,
                                     int version)
                              throws java.io.IOException

Throws:

java.io.IOException

estimatedMaxDepthForBytes

public static int estimatedMaxDepthForBytes(IPartitioner partitioner,
                                            long numBytes,
                                            int bytesPerHash)

Estimate the allowable depth while keeping the resulting heap usage of this tree under the provided number of bytes. This is important for ensuring that we do not allocate overly large trees that could OOM the JVM and cause instability. Calculated using the following logic: Let T = size of a tree of depth n T = #leafs * sizeof(leaf) + #inner * sizeof(inner) T = 2^n * L + 2^n - 1 * I T = 2^n * L + 2^n * I - I; So to solve for n given sizeof(tree_n) T: n = floor(log_2((T + I) / (L + I))

Parameters:

numBytes: - The number of bytes to fit the tree within

bytesPerHash: - The number of bytes stored in a leaf node, for example 2 * murmur128 will be 256 bits or 32 bytes

Returns:

the estimated depth that will fit within the provided number of bytes

get

public MerkleTree.TreeRange get(Token t)

For testing purposes. Gets the smallest range containing the token.