跳到主要导航 跳到搜索 跳到主要内容

PPMLT: A pipeline based processing model of long transactions

  • Yin Zhu*
  • , Feilong Tang
  • , Ilsun You
  • , Lin Lou
  • , Minyi Guo
  • , Yao Shen
  • *此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

In web service environment, long transactions always take long time to finish, which demands the resources -often the database - to be locked for a long time. This would bring down the performance of the transaction processing system. The normal solution is compensating transaction, which creates a relative transaction with the opposite effect. When the transaction is being executed, the sub-transactions are allowed to commit independently without waiting for other sub-transactions. If the transaction fails, the respective compensating transaction will be executed to eliminate the effect brought by the original transaction. However, there are still some problems in compensating transaction model. This paper proposes a novel pipeline based processing model of long transaction (PPMLT), which parallelizes the transaction processing mainly aimed at the serially executed long transaction. This model could improve the performance the transaction processing evidently without the compensating transaction. Our experiments show that the pipeline based processing model of serial long transaction has notable improvement when executing the long transaction.

源语言英语
主期刊名Proceedings - 25th IEEE International Conference on Advanced Information Networking and Applications, AINA 2011
200-207
页数8
DOI
出版状态已出版 - 2011
已对外发布
活动25th IEEE International Conference on Advanced Information Networking and Applications, AINA 2011 - Biopolis, 新加坡
期限: 22 3月 201125 3月 2011

出版系列

姓名Proceedings - International Conference on Advanced Information Networking and Applications, AINA
ISSN(印刷版)1550-445X

会议

会议25th IEEE International Conference on Advanced Information Networking and Applications, AINA 2011
国家/地区新加坡
Biopolis
时期22/03/1125/03/11

指纹

探究 'PPMLT: A pipeline based processing model of long transactions' 的科研主题。它们共同构成独一无二的指纹。

引用此