2012Unpublished venueRequires access

Resource Oriented Composite Transactions

Rui Liu, Xin Sheng Mao

Open publisher page 1 citations

Abstract

Composite transaction is a set of two or more related transactions managed in a coordinated logical unit of work. Composite transaction can span multiple data sources or multiple databases. Sometime, we use composite transaction only because we want to split long transactions into short transactions for less lock and more concurrency. In a very large scale distributed system, all of these are valid reasons to use composite transactions. There are different models for composite transaction. Distributed transaction model is widely used in practice since XA specification was released in 1991. However XA specification is notorious for heavy performance cost. To deal with internet scale transaction systems, we have to have some compromised solutions for both performance and consistency considerations. In this paper, we described the methodology and transaction programming model to simplify the system design and provide several guidelines for very large scale distributed transaction systems. We also described the differences between this design model and existing models for composite transaction including distributed transactions (XA 2PC), Sagas and federated transactions. Our design model is practicable and supports true concurrency. It targets at a special type of composite transactions named "Resource Oriented Composite Transaction" (ROCT) which is very common in banking systems, e-commerce systems, airline ticket-sale system, and etc. And our new transaction programming model can be widely used for this type of composite transactions.

About this research paper

What this paper is about

Composite transaction is a set of two or more related transactions managed in a coordinated logical unit of work. Composite transaction can span multiple data sources or multiple databases. Sometime, we use composite transaction only because we want to split long transactions into short transactions for less lock and more concurrency. In a very large scale distributed system, all of these are valid reasons to use composite transactions. There are different models for composite transaction. Distributed transaction model is widely used in practice since XA specification was released in 1991. However XA specification is notorious for heavy performance cost. To deal with internet scale transaction systems, we have to have some compromised solutions for both performance and consistency considerations. In this paper, we described the methodology and transaction programming model to simplify the system design and provide several guidelines for very large scale distributed transaction systems. We also described the differences between this design model and existing models for composite transaction including distributed transactions (XA 2PC), Sagas and federated transactions. Our design model is practicable and supports true concurrency. It targets at a special type of composite transactions named "Resource Oriented Composite Transaction" (ROCT) which is very common in banking systems, e-commerce systems, airline ticket-sale system, and etc. And our new transaction programming model can be widely used for this type of composite transactions.

Why it matters

OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

Composite transaction is a set of two or more related transactions managed in a coordinated logical unit of work. Composite transaction can span multiple data sources or multiple databases. Sometime, we use composite transaction only because we want to split long transactions into short transactions for less lock and more concurrency. In a very large scale distributed system, all of these are valid reasons to use composite transactions. There are different models for composite transaction. Distributed transaction model is widely used in practice since XA specification was released in 1991. However XA specification is notorious for heavy performance cost. To deal with internet scale transaction systems, we have to have some compromised solutions for both performance and consistency considerations. In this paper, we described the methodology and transaction programming model to simplify the system design and provide several guidelines for very large scale distributed transaction systems. We also described the differences between this design model and existing models for composite transaction including distributed transactions (XA 2PC), Sagas and federated transactions. Our design model is practicable and supports true concurrency. It targets at a special type of composite transactions named "Resource Oriented Composite Transaction" (ROCT) which is very common in banking systems, e-commerce systems, airline ticket-sale system, and etc. And our new transaction programming model can be widely used for this type of composite transactions.

Key concepts: Distributed transaction, Computer science, Transaction processing, Compensating transaction, Online transaction processing, Database transaction, Transaction processing system, Nested transaction

Related papers

Back to paper searchBrowse research topicsOriginal source
Resource Oriented Composite Transactions — Research Paper | ScholarLens