An approach to eliminate transaction blocking in locking protocols
An approach to eliminate transaction blocking in locking protocols
复制标题
一种消除锁定协议中事务阻塞的方法
DOI:
10.1145/137097.137877
复制
发表时间:
1992
期刊:
影响因子:
--
通讯作者:
Richard Jeffers
中科院分区:
文献类型:
--
作者:
D. Agrawal;A. E. Abbadi;Richard Jeffers
Locking based protocols are widely used for transaction management and concurrency control in database systems. Much of the recent wo~k on database concurrency control has concentrated on evaluating the performance of the two phase locking protocol [16] either by simulation studies [13, 33, 7, 5] or by analytical modeling [32, 17, 27 , 29, 34]. The simulation studies have generally established the superiority of two phase locking over other types of concurrency control protocols such as timest amp ordering [’28, 10] and optimistic concurrency control [22]. More recently several researchers have become interested in developing techniques to stabilize the behavior of two phase loc!iiug at very high multiprogramming levels (in particular beyond its thrash point). Tay, Suri, and Goodman [32] observed that if half the transactions ate blocked, the system is already thrashing. The authors then p~ovide a rule of thumb based on empirical observations that the workload should not exceed 1.5 to couttol thrashing. (Workload is a function It’N/D, where k is the transaction size, N is the multiprogramming level, and D is the database size.) Carey, Krishnamnrthi, and Livny [14] conducted extensive simulation studies and proposed the halj-and-hcdj Iule, which states that no mole than half the transactions shoulcl be blocked, to eliminate