• Background
  • The SkipTrie is a probabilistically-balanced version of a Y-Fast Trie. The Y-Fast Trie is the combination of an X-Fast Trie that's linked to a series of balanced binary search trees. The SkipTrie distinguishes itself from the Y-Fast Trie by forgoing the balanced binary search trees and replacing it with a single skiplist. By eliminating the need for rebalancing, the SkipTrie is essentially a lock-free linearizable implementation of the Y-fast trie.

• The Skip List

  • The skip list used in this project is Doug Lea's implementation of his skip list. We modified his skip list structure by adding list search and the coin fipping that is needed for the skip trie structure. We also modified the add method.

• The X-Fast Trie

  • Our implementation of the x-fast trie is essentially a pseudo-bitwise trie with a doubly linked list built on top of the leaf row in the trie. Each node has a concurrent hash map of its children.

• The Hash Table

  • The hashing thats done in the x-fast trie is done with Java's ConcurrentHashMap class.

• The Doubly Linked List

  • The doubly linked list is a direct port from Doug Lea's implementation of a concurrent doubly linked list.

• By combining all of the components in the Data Structures section we build an extremely fast concurrent search structure, the skip trie.
  • TODO: Description of how the search structure works

Joseph Leavitt, Joseph Landry, Jason Berk, Harold Marcial

• http://groups.csail.mit.edu/mag/oshman-shavit-podc13.pdf
• http://web.stanford.edu/class/archive/cs/cs166/cs166.1146/lectures/15/Small15.pdf
• https://en.wikipedia.org/wiki/Skip_list
• http://opendatastructures.org/versions/edition-0.1c/ods-java/node66.html