Two building blocks for replication: replicated abstract data types and $\texttt{log}^\prime$ version vector복제 추상화 데이터 구조와 로그 프라임 버전 벡터

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 607
  • Download : 0
Replication is a key technology that enables modern distributed applications to share data. This thesis suggests two building blocks that can facilitate various replication systems. First, to support efficient implementations of interactive collaborative applications, replicated abstract data types (RADTs) are proposed as building blocks. Collaborative applications highly desire responsive and transparent interactivity, and such interactivity can be achieved with optimistic replication. However, maintaining replica consistency is difficult. Concealing complicated consistency maintenance, RADTs replicate copies of a shared ADT and allow optimistic operations of different orders. In this thesis, a few representative abstract data types (ADTs), such as array, hash table, and growable array (or linked list), are extended into RADTs. Operation commutativity and precedence transitivity are two theoretical principles enabling RADTs to maintain consistency for optimistic operations of RADTs. Especially, precedence transitivity is a practical and efficient way to achieve consistency, for which an s4vector is proposed. Above all, replicated growable arrays (RGAs) can maintain consistency among variable-length ordered objects with insertion/deletion/update operations, which have been highly demanded and studied in collaborative applications for a long time. RGAs make significant improvement in complexity, scalability, and reliability over the effects of the previous approaches. In this dissertation, the experimental results show that the operations of RGAs are more than thousands of times faster than those of the operational transformation methods which have been mainstream approaches for optimistic insertion and deletion. Second, $\texttt{log}^\prime$ (log-prime) version vector is proposed for dynamic replication wherein replicas are frequently created and destroyed. In replication systems, version vectors are logged with replicas to detect conflicts among operations....
Advisors
Maeng, Seung-Ryoulresearcher맹승렬researcher
Description
한국과학기술원 : 전산학과,
Publisher
한국과학기술원
Issue Date
2011
Identifier
466470/325007  / 020037207
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전산학과, 2011.2, [ v, 56 p. ]

Keywords

version vector; consistency; distributed data structure; replication; logging overhead; 기록 오버헤드; 버전 벡터; 일관성; 분산 데이터 구조; 복제

URI
http://hdl.handle.net/10203/33330
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=466470&flag=dissertation
Appears in Collection
CS-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0