PPMLT: A Pipeline Based Processing Model of Long Transactions
Yin Zhu, Feilong Tang, Ilsun You, Lin Lou, Minyi Guo, Yao Shen
Abstract
Yin Zhu, Feilong Tang, Ilsun You, Lin Lou, Minyi Guo, Yao Shen
Abstract
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.
OpenAlex reports 3 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
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.
Key concepts: Transaction processing system, Transaction processing, Distributed transaction, Online transaction processing, Compensating transaction, Database transaction, Nested transaction, Computer science