Distributed hash table

From Wikipedia, the free encyclopedia

Distributed hash tables (DHTs) are a class of decentralized distributed systems that partition ownership of a set of keys among participating nodes, and can efficiently route messages to the unique owner of any given key. Each node is analogous to an array slot in a hash table. DHTs are typically designed to scale to large numbers of nodes and to handle continual node arrivals and failures. This infrastructure can be used to build more complex services, such as distributed file systems, peer-to-peer file sharing systems, cooperative web caching, multicast, anycast, domain name services, and instant messaging.

Contents

[edit] Background

DHT research was originally motivated, in part, by peer-to-peer systems such as Napster, Gnutella, and Freenet, which took advantage of resources distributed across the Internet to provide a single useful application. In particular, they took advantage of increased bandwidth and hard disk capacity to provide a file sharing service.

These systems differed in how they found the data their peers contained. Napster had a central index server: each node, upon joining, would send a list of locally held files to the server, which would perform searches and refer the querier to the nodes that held the results. This central component left the system vulnerable to attacks and lawsuits. Gnutella and similar networks moved to a flooding query model — in essence, each search would result in a message being broadcast to every other machine in the network. While avoiding a single point of failure, this method was significantly less efficient than Napster. Finally, Freenet was also fully distributed, but employed a heuristic key based routing in which each file was associated with a key, and files with similar keys tended to cluster on a similar set of nodes. Queries were likely to be routed through the network to such a cluster without needing to visit many peers. However, Freenet did not guarantee that data would be found.

Distributed hash tables use a more structured key based routing in order to attain both the decentralization of Gnutella and Freenet, and the efficiency and guaranteed results of Napster. One drawback is that, like Freenet, DHTs only directly support exact-match search, rather than keyword search, although that functionality can be layered on top of a DHT.

The first four DHTs—CAN, Chord, Pastry, and Tapestry—were introduced about the same time in 2001. Since then this area of research has been quite active. Outside academia, DHT technology has been adopted as a component of BitTorrent and in the Coral Content Distribution Network.

[edit] Properties

DHTs characteristically emphasize the following properties:

  • Decentralisation: the nodes collectively form the system without any central coordination.
  • Scalability: the system should function efficiently even with thousands or millions of nodes.
  • Fault tolerance: the system should be reliable (in some sense) even with nodes continuously joining, leaving, and failing.

A key technique used to achieve these goals is that any one node needs to coordinate with only a few other nodes in the system – most commonly, Θ(logn) of the n participants (see below) – so that only a limited amount of work needs to be done for each change in membership.

Some DHT designs seek to be secure against malicious participants and to allow participants to remain anonymous, though this is less common than in many other peer-to-peer (especially file sharing) systems; see anonymous P2P.

Finally, DHTs must deal with more traditional distributed systems issues such as load balance, data integrity, and performance (in particular, ensuring that operations such as routing and data storage or retrieval complete quickly).

[edit] Structure

A DHT is built around an abstract keyspace, such as the set of 160-bit strings. Ownership of the keyspace is split among the participating nodes according to a keyspace partitioning scheme. The overlay network connects the nodes, allowing them to find the owner of any given key in the keyspace. (This design decomposition has been suggested in (Naor and Wieder, 2003) and (Manku, 2004).)

Once these components are in place, a typical use of the DHT for storage and retrieval might proceed as follows. Suppose the keyspace is the set of 160-bit strings. To store a file with given filename and data in the DHT, the SHA1 hash of filename is found, producing a 160-bit key k. Thereafter, a message put(k,data) may be sent to any node participating in the DHT. The message is forwarded from node to node through the overlay network until it reaches the single node responsible for key k as specified by the keyspace partitioning, where the pair (k,data) is stored. Any other client can then retrieve the contents of the file by again hashing filename to produce k and asking any DHT node to find the data associated with k with a message get(k). The message will again be routed through the overlay to the node responsible for k, which will reply with the stored data.

Each of these components is described below with the goal of capturing the principal ideas common to most DHTs; many designs differ in the details.

[edit] Keyspace partitioning

Most DHTs use some variant of consistent hashing to map keys to nodes. This technique employs a function δ(k1,k2) which defines an abstract notion of the distance from key k1 to key k2. Each node is assigned a single key called its identifier (ID). A node with ID i owns all the keys for which i is the closest ID, measured according to δ.

Example. The Chord DHT treats keys as points on a circle, and δ(k1,k2) is the distance traveling clockwise around the circle from k1 to k2. Thus, the circular keyspace is split into contiguous segments whose endpoints are the node identifiers. If i1 and i2 are two adjacent IDs, then the node with ID i2 owns all the keys that fall between i1 and i2.

Consistent hashing has the essential property that removal or addition of one node changes only the set of keys owned by the nodes with adjacent IDs, and leaves all other nodes unaffected. Contrast this with a traditional hash table in which addition or removal of one bucket causes nearly the entire keyspace to be remapped. Since any change in ownership typically corresponds to bandwidth-intensive movement of objects stored in the DHT from one node to another, minimizing such reorganization is required to efficiently support high rates of churn (node arrival and failure).

[edit] Overlay network

Each node maintains a set of links to other nodes (its neighbors or routing table). Together these form the overlay network, and are picked in a structured way, called the network's topology.

All DHT topologies share some variant of the most essential property: for any key k, the node either owns k or has a link to a node that is closer to k in terms of the keyspace distance defined above. It is then easy to route a message to the owner of any key k using the following greedy algorithm: at each step, forward the message to the neighbor whose ID is closest to k. When there is no such neighbor, then we must have arrived at the closest node, which is the owner of k as defined above. This style of routing is sometimes called key based routing.

Beyond basic routing correctness, two key constraints on the topology are to guarantee that the maximum number of hops in any route (route length) is low, so that requests complete quickly; and that the maximum number of neighbors of any node (maximum node degree) is low, so that maintenance overhead is not excessive. Of course, having shorter routes requires higher maximum degree. Some common choices for maximum degree and route length are as follows, where n is the number of nodes in the DHT:

  • Degree O(1), route length O(logn)
  • Degree O(logn), route length O(logn / loglogn)
  • Degree O(logn), route length O(logn)
  • Degree O(n1 / 2), route length O(1)

The third choice is the most common, even though it is not quite optimal in terms of degree/route length tradeoff, because such topologies typically allow more flexibility in choice of neighbors. Many DHTs use that flexibility to pick neighbors which are close in terms of latency in the physical underlying network.

Maximum route length is closely related to diameter: the maximum number of hops in any shortest path between nodes. Clearly the network's route length is at least as large as its diameter, so DHTs are limited by the degree/diameter tradeoff [1] which is fundamental in graph theory. Route length can be greater than diameter since the greedy routing algorithm may not find shortest paths (Manku et al 2004).

[edit] Examples

[edit] DHT protocols and implementations

[edit] Applications employing DHTs

[edit] See also

  • memcached: a high-performance, distributed memory object caching system
  • Tangosol Coherence includes a structure similar to a DHT, though all nodes have knowledge of the other participants

[edit] Articles

[edit] External links