Conference Paper

Efficient replication control in distributed real-time databases

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Summary form only given. Recently, the demand for real-time data services has been increasing. Many e-commerce applications and information services are requiring sophisticated real-time data support. A database is a core component for such real-time applications. Important to the functionality of real-time databases is data replication, which is used to meet critical time requirements. These requirements vary with different workloads; therefore, different methods of replication control are better suited for different workloads. We examine several methods of replication control and present two new methods for distributed real-time database systems. Our methods are more ideally suited for systems with non-static periodic transaction arrival patterns and systems with random transaction arrival patterns.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... A real-time database system (RTDBS) is defined in [2] as a database system that includes all features of traditional database system, while enforcing real-time constraints or deadlines. According to [3], the time constraints can be on the data level in the form of time validation attribute making temporal data whose validity is lost after the elapse of some pre_specified time interval, or on the transaction level in the form of deadline used by the real-time scheduling and concurrency control. ...
... Clustering is a process of partitioning a given set of objects into groups of similar objects based on similarity metrics. Clustering analysis is broadly used in distributed systems researches specially to segment the network sites based on many similarity metrics to reduce the traffic load [2]. The resulting cluster can be treated as one group which is considered as a form of data compression [7]. ...
... Many clustering techniques have been introduced in a wide variety of applications such as image processing [2], network segmentation [13], marketing [14], pattern recognition [15], and customer segmentation [10]. ...
... A real-time database system (RTDBS) is a time constraint system in which a transaction requires servicing on or before its deadline (Aslinger & Sang 2005). The resulting value from missing the deadline is used to categorize the real-time transactions into three types, soft, firm, and hard. ...
... Although there are many researchers working in the field of DRTDBS replication such as: (Aslinger & Sang 2005); (Li-Wei & Yang 2011); (Hamdi, Salem & Bouazizi 2013); and Said, Sadeg & Ayeb 2009), few of them address the concept of virtual full replication such as described in (Mathiason, Andler & Son 2007) and ( Gustafson et al. 2004). The work by (Mathiason, Andler & Son 2007) tries to maintain the scalability at execution time and to improve the access time to data at any node by grouping data objects into segments. ...
... Others works target minimization of the execution duration of the updating operation including work by (Aslinger & Sang 2005); (Said, Sadeg & Ayeb 2009); and (Xiong et al.2010). The idea of reducing the execution time of this category of work is close to our vision. ...
Article
Full-text available
Scalability is an increasingly important target for distributed real-time databases. Replication is widely applied to improve the scalability and availability of data. With full replication, database systems cannot scale well, since all updates must be replicated to all nodes, whether or not they are needed there. With virtual Full Replication, all nodes have an image of a fully replicated database and the system manages the knowledge of what is needed for each node to adapt to the actual needs, so that the system can be more scalable. This work proposes a scalable and consistent replication protocol using an adaptive clustering technique that dynamically detects the new data requirements. Because time is critical in such systems, the clustering technique must take into account both the communication time cost and the timing properties of the data. The proposed protocol also proposes a new updated method for addressing the temporal inconsistency problem by skipping unnecessary operations. It allows many database nodes to update their data concurrently, without any need for distributed synchronization. It uses state-transfer propagation with on-demand integration techniques to reduce the temporal inconsistency. The experimental results show the ability of the proposed protocol to reduce the system resources consumed and improves system scalability while maintaining consistency.
... A real-time database system (RTDBS) is defined in [2] as a database system that includes all features of traditional database system, while enforcing real-time constraints or deadlines. According to [3], the time constraints can be on the data level in the form of time validation attribute making temporal data whose validity is lost after the elapse of some pre_specified time interval, or on the transaction level in the form of deadline used by the real-time scheduling and concurrency control. ...
... Clustering is a process of partitioning a given set of objects into groups of similar objects based on similarity metrics. Clustering analysis is broadly used in distributed systems researches specially to segment the network sites based on many similarity metrics to reduce the traffic load [2]. The resulting cluster can be treated as one group which is considered as a form of data compression [7]. ...
... II. RELATED WORK Many clustering techniques have been introduced in a wide variety of applications such as image processing [2], network segmentation [13], marketing [14], pattern recognition [15], and customer segmentation [10]. ...
Article
Recently, the demand for real-time database is increasing. Most real-time systems are inherently distributed in nature. They need data to be obtained and updated in a timely fashion. Sometimes required data are at a particular location is not available, and needed to be obtained from remote site. This may take long time that make the temporal data invalid resulting in large number of tardy transactions with their fetal effect. Clustering the database sites nodes can help distributed real-time database systems to face the challenges meeting their time requirements. Reducing the large number of network sites into many clusters with smaller number of sites will effectively decrease the response time, resulting in better meeting of time constraints. In this work, we introduce a novel clustering algorithm for distributed real-time database that depend on both the communication time cost and the timing properties of data. The results showed lower communication time, higher database performance and better meeting of timing requirements.
... A Real-Time Database System (RTDBS) is defined in [3] as a database system that includes all features on traditional database system, in addition to enforcing time-constraints in a form of data validation duration or transaction deadlines or both. According to [19,20], the result value from missing deadline is used to categorize the real-time transactions into three types, soft, firm and hard. ...
... Predictability is the most important real-time feature of RTDBS. It is often more important than consistency, which can be relaxed to improve predictability of data accesses [3]. That is due to the fatal effect of missing deadline in a hard real-time systems and the reduced service of soft real-time systems [21]. ...
Article
Full-text available
Recently, the demand for real-time database is increasing. Most real-time systems are inherently distributed in nature and need to handle data in a timely fashion. Obtaining data from remote sites may take long time making the temporal data invalid. This results in a large number of tardy transactions with their catastrophic effect. Replication is one solution of this problem, as it allows transactions to access temporal data locally. This helps transactions to meet their time requirements which require predictable resource usage. To improve predictability, DeeDS prototype is introduced to avoid the delay which results from disk access, network communications and distributed commit processing. DeeDS advises to use In-memory database, fully replication and local transaction committing, but full replication consumes the system resources causing a scalability problem. In this work, we introduce IReIDe as a new replication protocol that supports the replication for DeeDS and faces the scalability problem using intelligent clustering technique. The results show the ability of IReIDe to reduce the consumed system resources and maintain scalability.
... This means if p i is does not crash itself by end of round r + 1 + 2R, then at least f + 1 correct processes have received and signed p i s Deliver(m) message by round r + 2R. By lines (42)(43)(44)(45)(46)(47)(48)(49)(50)(51) of Algorithm 4 all such correct processes deliver m at latest by round r + 2R. ...
Article
Full-text available
Today's cyber-physical systems face various impediments in achieving their intended goals. Namely, communication uncertainties and faults, relative to the increased integration of networked and wireless devices, hinder the synchronism needed to meet real-time deadlines. Moreover, being critical, these systems have higher security threats that cause physical damage. This paper addresses the aforementioned problems by studying how to build the first real-time Byzantine reliable broadcast (RTBRB) protocol tolerating network uncertainties, faults, and attacks. Previous literature contains either real-time reliable broadcast protocols, or asynchronous (non real-time) Byzantine ones.
Preprint
Full-text available
Today's cyber-physical systems face various impediments to achieving their intended goals, namely, communication uncertainties and faults, relative to the increased integration of networked and wireless devices, hinder the synchronism needed to meet real-time deadlines. Moreover, being critical, these systems are also exposed to significant security threats. This threat combination increases the risk of physical damage. This paper addresses these problems by studying how to build the first real-time Byzantine reliable broadcast protocol (RTBRB) tolerating network uncertainties, faults, and attacks. Previous literature describes either real-time reliable broadcast protocols, or asynchronous (non real-time) Byzantine~ones. We first prove that it is impossible to implement RTBRB using traditional distributed computing paradigms, e.g., where the error/failure detection mechanisms of processes are decoupled from the broadcast algorithm itself, even with the help of the most powerful failure detectors. We circumvent this impossibility by proposing RT-ByzCast, an algorithm based on aggregating digital signatures in a sliding time-window and on empowering processes with self-crashing capabilities to mask and bound losses. We show that RT-ByzCast (i) operates in real-time by proving that messages broadcast by correct processes are delivered within a known bounded delay, and (ii) is reliable by demonstrating that correct processes using our algorithm crash themselves with a negligible probability, even with message loss rates as high as 60%.
Chapter
Replication is the technique to provide the opportunity of deadline fulfillment of real-time transaction in distributed real-time database system (DRTDBS). Just-in-time real-time replication (JITRTR), on-demand real-time replication (ORDER), on-demand real-time replication with replica sharing (ORDER-RS), periodic transactions using partial replication (PTUDPR), periodicity prediction algorithm, and so on are the replication protocols designed for DRTDBS. None of the algorithms follow one-copy serializability (1SR) correctness criteria for real-time and non-real-time data items. Our main objective is to fulfill strict consistency together with transaction deadline fulfillment. In this paper, the proposed model works on multi-master together with slave sites and uses middleware for following 1SR correctness criteria. In the proposed model, master sites are responsible for executing update and write transactions and slave sides are responsible for executing only read transaction. The proposed model performs better in terms of deadline fulfillment and strict consistency with respect to other replication protocol.
Article
Full-text available
There is a growing need for real-time data services in distributed environments. Providing quality-of-service guarantees for data services in a distributed environment is a challenging task. The presence of multiple sites in distributed environments raises issues that are not present in cen- tralized systems. The transaction workloads in distributed real-time databases may not be balanced and the transaction access patterns may be time-varying and skewed. Data replication is an eective method to help database systems meet the stringent temporal requirements of real-time applica- tions. We have designed an algorithm that provides quality-of-service guarantees for data services in distributed real-time databases with full replication of temporal data. The algorithm consists of heuristic feedback-based local controllers and global load balancers (GLB) working at each site. The local controller controls the admission process of incoming transactions. The global load balancers collect the performance data from other nodes and balance the system-wide workload. The simula- tion results show that the new algorithm successfully balances the workloads in distributed real-time databases and provides tight transaction miss ratio guarantees under various transaction workloads.
Conference Paper
Full-text available
Providing quality of service guarantees for data services in a distributed environment is a challenging task. The presence of multiple sites in distributed environment raises issues that are not present in centralized systems. The transaction workloads in distributed real-time databases may not be balanced and the transaction access patterns may be time-varying and skewed. Data replication is an effective method to help database systems meet the stringent temporal requirements of real-time applications. We have designed an algorithm that provides quality of service guarantees for data services in distributed real-time databases with full replication of temporal data. The algorithm consists of heuristic feedback-based local controllers and global load balancers (GLB) working at each sites. The local controller controls the admission process of incoming transactions. The global load balancers collect the performance data from other nodes and balance the system-wide workload. The simulation results show that the new algorithm successfully balances the workloads in distributed real-time databases and provides tight transaction miss ratio guarantees under various transaction workloads.
Conference Paper
Full-text available
The demand for real-time database services has been increasing recently. Examples include sensor data fusion, decision support, Web information services, and online trading. In these applications, it is desirable to execute transactions within their deadlines using temporally consistent data. Due to the high service demand, real-time databases can be overloaded. As a result, many transactions may miss their deadlines, or data temporal consistency constraints can be violated. To address these problems, we present a QoS management scheme to support guarantees on deadline miss ratio and data freshness (temporal consistency) even in the presence of unpredictable workloads and data access patterns. Using our approach, admitted user transactions can be processed in time using fresh data. A simulation study shows that our QoS-sensitive approach can achieve a significant performance improvement, in terms of deadline miss ratio and data freshness, compared to several baseline approaches. Furthermore, our approach shows a comparable performance to the theoretical oracle that is privileged by a complete future knowledge of data accesses.
Article
Full-text available
The demand for real-time data services is increasing in many applications including e-commerce, agile manufacturing, and telecommunications network management. In these applications, it is desirable to execute transactions within their deadlines, i.e., before the real-world status changes, using fresh (temporally consistent) data. However, meeting these fundamental requirements is challenging due to dynamic workloads and data access patterns in these applications. Further, transaction timeliness and data freshness requirements may conflict. We define average/transient deadline miss ratio and new data freshness metrics to let a database administrator specify the desired quality of real-time data services for a specific application. We also present a novel QoS management architecture for real-time databases to support the desired QoS even in the presence of unpredictable workloads and access patterns. To prevent overload and support the desired QoS, the presented architecture applies feedback control, admission control, and flexible freshness management schemes. A simulation study shows that our QoS-aware approach can achieve a near zero miss ratio and perfect freshness, meeting basic requirements for real-time transaction processing. In contrast, baseline approaches fail to support the desired miss ratio and/or freshness in the presence of unpredictable workloads and data access patterns.
Article
Full-text available
The demand for real-time database services has been increasing recently. Examples include sensor data fusion, decision support, web information services, e-commerce, online trading, and data-intensive smart spaces. A real-time database, a core component of many information systems for real-time applications, can be a main service bottleneck. To address this problem, we present a real-time database QoS management scheme that provides guarantees on deadline miss ratio and data freshness, which are considered two fundamental performance metrics for real-time database services. Using our approach, admitted user transactions can be processed in a timely manner and data freshness can be guaranteed even in the presence of unpredictable workloads and data access patterns. A simulation study shows that our QoS-sensitive approach can achieve a significant performance improvement, in terms of deadline miss ratio and data freshness, compared to several baseline approaches. Furthermore, our approach shows a comparable performance to the theoretical oracle that is privileged by the complete future knowledge of data accesses.
Chapter
These four sample application domains, while certainly not representative of all real-time applications potentially using databases, illustrate some of the widely diverse requirements involved. It is interesting to note that the numerical values for the application data characteristics shown for each domain are gen- erally stated neither in the “official” application requirements documentation, nor in the design documentation that reflects derived application requirements. These figures are not described in system requirements or design because it is not generally recognized that these characteristics drive the design of the data-handling mechanisms to be used by the application. Instead, most actual application designs either use ad-hoc data handling mechanisms, or attempt to use normal commerical database products that provide either anomalous cosistency behavior, or inappropriate response characteristics due to excessive adherence to the ACID properties. One of the main purposes of this book is to demonstrate how application designers can use these derived data requirements to define appropriate characteristics of real-time databases that will support them.
Conference Paper
In this edited book, we divide all important design issues of RTDBS into six parts. Part I is focused on the basic concepts of RTDBS and the clarification their mis-concepts. Part II discussed one of the most important issues in the design of RTDBS: concurrency control. Other issues related to the management of the systems are included in Part III. Part IV addresses real-time active issues, such as triggered transactions. Part V reports the research findings in distributed and mobile real-time database systems. Part VI is for prototype and future work.
Conference Paper
This paper describes a set of characteristics that partially make up the problem space for real-time data distribution. The characteristics are divided into three groups: system characteristics, real-time characteristics, and data characteristics. The paper describes each of the characteristics, and then illustrates them with two current proposed solutions to real-time data distribution problem.
Conference Paper
This paper describes a replication algorithm for distributed real-time object-oriented databases in a static environment. All data requirements are specified a priori, and the algorithm creates replication transactions that copy remote data to a local site in order to guarantee that even, data request reads temporally valid data. The algorithm conditions are proven to be necessary and sufficient for providing this guarantee. Test results indicate that under most conditions, this replication strategy is better than total replication, which is a typical strategy used in distributed databases
New Features in AIM 5.2.3292
  • Aol Instant
  • Messenger
TimesTen White Paper Available in the World Wide Web
  • Timesten Performance Software