Title :
Materialized views for eventually consistent record stores
Author :
Changjiu Jin ; Rui Liu ; Salem, Kenneth
Author_Institution :
Cheriton Sch. of Comput. Sci., Univ. of Waterloo, Waterloo, ON, Canada
Abstract :
Distributed, replicated keyed-record stores are often used by applications that place a premium on high availability and scalability. Such systems provide fast access to stored records given a primary key value, but access without the primary key may be very slow and expensive. This problem can be addressed using materialized views. Materialized views redundantly store records, or parts of records, and the redundant copies can be organized and distributed differently than the originals, e.g, according to the value of a secondary key. In this paper, we consider the problem of supporting materialized views in multi-master, eventually consistent keyed-record stores. Incremental maintenance of materialized views is challenging in such systems because there no single master server responsible for serializing the updates to each record. We present a decentralized technique for incrementally maintaining materialized views in multi-master systems. We have implemented a prototype of our technique using Cassandra, a widely used system of this type. Using the prototype, we show that secondary-key-based access is much faster using materialized views than using Cassandra´s native secondary indexes, but maintaining the views in the face of updates may be more expensive than maintaining indexes.
Keywords :
database indexing; replicated databases; Cassandra native secondary index; decentralized technique; distributed replicated keyed-record store; eventually consistent keyed-record stores; eventually consistent record store; index maintenance; materialized view; multimaster system; primary key value; redundant copy distribution; redundant copy organization; redundantly store records; secondary-key-based access; stored record access; update serialization; Availability; Educational institutions; Indexes; Maintenance engineering; Scalability; Servers;
Conference_Titel :
Data Engineering Workshops (ICDEW), 2013 IEEE 29th International Conference on
Conference_Location :
Brisbane, QLD
Print_ISBN :
978-1-4673-5303-8
Electronic_ISBN :
978-1-4673-5302-1
DOI :
10.1109/ICDEW.2013.6547459