M.Sc Thesis

M.Sc StudentSheffi Gal
SubjectA Scalable Linearizable Multi-Index Table
DepartmentDepartment of Computer Science
Supervisor PROF. Erez Petrank
Full Thesis textFull thesis text - English Version


Concurrent data structures typically index data using a single primary key and provide fast access to data associated with a given key value. However, it is often required to access information via multiple primary and secondary keys, and even through additional properties that do not represent keys for the given data. One way to implement a data structure that allows access via multiple indexing is to use an existing data structure that allows access via a single index, and traverse the items exhaustively to find items by a secondary key. However, using this method, it is difficult to achieve either correctness (linearizability) or good performance. In this thesis, we propose to obtain multi-indexing access by using an efficient concurrent data structure for each of the indexed fields. We then coordinate these structures to obtain correctness and good performance. We propose lock-free and lock-based designs of a table with multiple indexing, supporting linearizable inserts, deletes, and retrieve operations. We have implemented the proposed designs and ran measurements on a 64-core AMD platform. The obtained performance demonstrates efficiency and high scalability. In particular, measurements show that lock-free solutions perform better than lock-based ones when contention is high, as well as in most practical scenarios.