Performance Analysis of Concurrency Control Mechanisms for OLTP Databases

Abstract

Abstract-Concurrency control (CC) in distributed and multidimensional databases is becoming more important due to recent increase in high-volume data storage with increasing online transaction processing (OLTP) requirements for medium and large organisations. This paper examines three concurrency control mechanisms commonly adopted and analyses their performance in distributed databases for OLTP operational systems of enterprises. The three CC mechanisms investigated are, two phase locking (2PL), wait depth limited (WDL) and optimistic concurrency control. These CC mechanisms have been studied well in disk-based systems. However, with the recent advances of cost-effective main memory or in-memory storage that can support much higher transaction rates than disk-based systems, there is sufficient motivation to re-investigate the performance of such CC mechanisms in diverse processor configurations. This paper presents a comparison of their behaviour and performance in terms of throughput rates achieved with varying transaction size and contention. The outcome of this study has resulted in further research proposals for improving the performance of these CC mechanisms for OLTP databases.Index Terms-Concurrency control, online transaction processing (OLTP), in-memory databases, performance.

Performance comparison of in-memory and disk-based databases using transaction processing performance council (TPC) benchmarking

“…Performance evaluation of database requires comprehensive defined tests to measure two main areas, throughput and response time (Kaspi and Venkatraman, 2014;TPC, 2010). In terms of comparison of disk-based and in-memory table design, it is important to have either same hardware or identical servers with same internal configuration and parameters.…”

“…Different organizations have different workloads; even within a particular organization, these workloads represent various statistics to measure and benchmark for future cross verifications. In general, the following are the most popular database workload types (Kaspi and Venkatraman, 2014;Elnaffar et al, 2002):…”

“…Online transactional processing based application requires comprehensive planning prior to deployment in regards to database performance and throughput. The ideal goal for the database system is to perform each transaction in shortest possible response time for application-specific structured query language (SQL) queries (Kaspi and Venkatraman, 2014;TPC, 2010). In most cases of online transaction processing (OLTP), concurrent users of the specific module of the application can cause congestion and increase resource locking (Faleiro and Abadi, 2011).…”

Survey on Concurrency Control Techniques

“…[12] This protocol ensures conflictserializable schedules. Such described in [1], [2], [3], [14][15].…”

“…In optimistic concurrency control, that is also called validation based technique, check is done after read phase and before write phase to minimize overhead during transaction execution, with the effect of slowing down the transactions. Here transaction passes with three phases, read phase in which transaction can read committed data from database, validation phase that check whether update done by transaction will not affect database consistency, write phase if validation success transaction updates are applied to database else updates will be discarded and transaction is restarted, such as [2], [3], [5], [6], [13], [14], and [19].…”