Publicação

Improving Lazy State Determination

Ver documento

Detalhes bibliográficos
Resumo:High-performance transaction systems are used for online transaction processing in major platforms, like online shopping or social media platforms. It is important that these systems exhibit high throughput, with a low-latency, so that the final users do not experience unresponsiveness while browsing the platform. When under high contention, these systems often experience high abort rates or deadlocks, which is not ideal for the use-cases these systems are made for. Lazy State Determination (LSD) helps by delaying, as much as possible, the need to retrieve concrete data from the database system, thus decreasing the window of opportunity for the data to be changed by other transactions, which improves transaction success rates. In order to create and execute transactions, LSD uses futures instead of concrete values, proving operations to compose these futures. When committing a transaction, LSD evaluates all futures by fetching concrete values, completing the transaction successfully if the conditions are maintained. We propose a solution to improve the shortcomings of the LSD specification, as presented by Vale [25] and Carpinteiro [3].
Autores principais:Margalhau, Ricardo Alexandre Cruz
Assunto:Transactions Database Management Systems Lazy State Determination High Contention Concurrency Management High-performance Transactions
Ano:2025
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:High-performance transaction systems are used for online transaction processing in major platforms, like online shopping or social media platforms. It is important that these systems exhibit high throughput, with a low-latency, so that the final users do not experience unresponsiveness while browsing the platform. When under high contention, these systems often experience high abort rates or deadlocks, which is not ideal for the use-cases these systems are made for. Lazy State Determination (LSD) helps by delaying, as much as possible, the need to retrieve concrete data from the database system, thus decreasing the window of opportunity for the data to be changed by other transactions, which improves transaction success rates. In order to create and execute transactions, LSD uses futures instead of concrete values, proving operations to compose these futures. When committing a transaction, LSD evaluates all futures by fetching concrete values, completing the transaction successfully if the conditions are maintained. We propose a solution to improve the shortcomings of the LSD specification, as presented by Vale [25] and Carpinteiro [3].