Multi-granularity Locking in Hierarchies with Synergistic Hierarchical and Fine-Grained Locks

被引:4
作者
Ganesh, K. [1 ]
Kalikar, Saurabh [1 ]
Nasre, Rupesh [1 ]
机构
[1] IIT Madras, CSE, Chennai, Tamil Nadu, India
来源
EURO-PAR 2018: PARALLEL PROCESSING | 2018年 / 11014卷
关键词
D O I
10.1007/978-3-319-96983-1_39
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
We propose a new locking mechanism for hierarchies wherein the locking requests can be a combination of coarse and fine. Existing protocols such as multiple-granularity locking (MGL) are efficient when all the requests are of the same granularity. MGL is either too coarse or too fine-grained when multiple threads request for various parts of the hierarchy with differing granularity requirements. Simultaneous handling of hierarchical and fine-grained requests poses new challenges in checking for racy requests. We propose a novel indexing technique for hierarchies which uniquely identifies every node as an interval value and effectively captures hierarchical dependencies between nodes even when the hierarchy is a tree, DAG or a cycle. Our experiments with real-world XML hierarchies and synthetic benchmarks show that the proposed locking technique provides a higher degree of concurrency with minimal locking cost resulting in overall performance improvement.
引用
收藏
页码:546 / 559
页数:14
相关论文
共 16 条
[11]  
Lomet D, 2009, PROC VLDB ENDOW, V2
[12]   Autolocker: Synchronization inference for atomic sections [J].
McCloskey, B ;
Zhou, F ;
Gay, D ;
Brewer, E .
ACM SIGPLAN NOTICES, 2006, 41 (01) :346-358
[13]  
Ries D. R., 1979, ACM Transactions on Database Systems, V4, P210, DOI 10.1145/320071.320078
[14]  
Ries D. R., 1977, ACM Transactions on Database Systems, V2, P233, DOI 10.1145/320557.320566
[15]  
Treebank, 2002, XML DAT REP
[16]  
Unrau R. C., 1994, Proceedings of the First USENIX Symposium on Operating Systems Design and Implementation (OSDI), P139