Paul M. Bird

Extending relational database systems to automatically enforce privacy policies

Databases are at the core of successful businesses. Due to the voluminous stores of personal data being held by companies today, preserving privacy has become a crucial requirement for operating a business. This paper proposes how current relational database management systems can be transformed into their privacy-preserving equivalents. Specifically, we present language constructs and implementation design for fine-grained access control to achieve this goal.

Workload adaptation in autonomic DBMSs

Workload adaptation is a performance management process in which an autonomic database management system (DBMS) efficiently makes use of its resources by filtering or controlling the workload presented to it in order to meet its Service Level Objectives (SLOs). This paper presents a framework and a prototype implementation of a query scheduler that performs workload adaptation in a DBMS. The system manages multiple classes of queries to meet their performance goals by allocating DBMS resources through admission control in the presence of workload fluctuation. The resource allocation plan is derived by maximizing the objective function that encapsulates the performance goals of all classes and their importance to the business. A first-principle performance model is used to predict the performance under the new resource allocation plan. Experiments with IBM® DB2® Universal Database#8482; are conducted to show the effectiveness of the framework.

A multi-purpose implementation of mandatory access control in relational database management systems

Mandatory Access Control (MAC) implementations in Relational Database Management Systems (RDBMS) have focused solely on Multilevel Security (MLS). MLS has posed a number of challenging problems to the database research community, and there has been an abundance of research work to address those problems. Unfortunately, the use of MLS RDBMS has been restricted to a few government organizations where MLS is of paramount importance such as the intelligence community and the Department of Defense. The implication of this is that the investment of building an MLS RDBMS cannot be leveraged to serve the needs of application domains where there is a desire to control access to objects based on the label associated with that object and the label associated with the subject accessing that object, but where the label access rules and the label structure do not necessarily match the MLS two security rules and the MLS label structure. This paper introduces a flexible and generic implementation of MAC in RDBMS that can be used to address the requirements from a variety of application domains, as well as to allow an RDBMS to efficiently take part in an end-to-end MAC enterprise solution. The paper also discusses the extensions made to the SQL compiler component of an RDBMS to incorporate the label access rules in the access plan it generates for an SQL query, and to prevent unauthorized leakage of data that could occur as a result of traditional optimization techniques performed by SQL compilers.

Autonomic workload execution control using throttling

Database Management Systems (DBMSs) are often required to simultaneously process multiple diverse workloads while enforcing business policies that govern workload performance. Workload control mechanisms such as admission control, query scheduling, and workload execution control serve to ensure that such policies are enforced and that individual workload goals are met. Query throttling can be used as a workload execution control method whereby problematic queries are slowed down, thus freeing resources to allow the more important work to complete more rapidly. In a self-managed system, a controller would be used to determine the appropriate level of throttling necessary to allow the important workload to meet is goals. The throttling would be increased or decreased depending upon the current system performance. In this paper, we explore two techniques to maintain an appropriate level of query throttling. The first technique uses a simple controller based on a diminishing step function to determine the amount of throttling. The second technique adopts a control theory approach that uses a black-box modelling technique to model the system and to determine the appropriate throttle value given current performance. We present a set of experiments that illustrate the effectiveness of each controller, then propose and evaluate a hybrid controller that combines the two techniques.

DBMS workload control using throttling: experimental insights

Todays database management systems (DBMSs) are required to handle diverse, mixed workloads and to provide differentiated levels of service to ensure that critical work takes priority. In order to meet these needs, it is necessary for a DBMS to have control over the workload executing in the system. Lower priority workloads should be limited to allow higher priority workloads to complete in a timely fashion. In this paper we examine query throttling techniques as a method of workload control. In our approach, a workload class may be slowed down during execution in order to release system resources that can be used by higher priority workloads. We examine two methods of throttling; constant throttling throughout query execution, and a single interruption in which a query is paused for a period of time. A set of experiments using Postresql 8.1 provides insights regarding the performance of these different throttling techniques under different workload conditions and how they compare to using operating system process priority control as a throttling mechanism.

Using economic models to allocate resources in database management systems

Resource allocation in database management systems is a performance management process in which an autonomic DBMS makes resource allocation decisions based on properties like workload business importance. We propose the use of economic models to guide the resource allocation decisions. An economic model is described in terms of business concepts and has been successfully applied in computer system resource allocation problems. In this paper, we present an approach that uses economic models to allocate multiple resources, such as main memory buffer space and CPU shares, to workloads running concurrently on a DBMS. The economic model enables workloads to meet their service level objectives by allocating resources through partitioning the individual DBMS resources and making system-level resource allocation plans for the workloads. The resource allocation plans can be dynamically changed to respond to changes in workload performance requirements. Experiments are conducted on IBM® DB2® databases to verify the effectiveness of our approach.

Poster Session: Adapting Mixed Workloads to Meet SLOs in Autonomic DBMSs

Workload adaptation allows an autonomic database management system (DBMS) to efficiently make use of its resources and meet its service level objectives (SLOs) by filtering or controlling the workload presented to it. Workload adaptation has been shown to be effective for OLAP and OLTP workloads. We outline a framework of workload adaptation and explain how it can be extended to manage mixed workloads comprised of both OLAP and OLTP queries. Experiments with IBM DB2 Universal Database are presented that illustrate the effectiveness of our techniques.

Discovering Indicators for Congestion in DBMSs

In todays data server environments, multiple types of workloads can be present in a system simultaneously. Workloads may have different levels of business importance and unique performance goals. An autonomic workload management system controls the flow of the workloads to help the database management system (DBMS) meet the performance goals. A task of the autonomic workload management system is to prevent congestion in the DBMS, which can result in severe degradation in overall system performance. Autonomic workload management should detect that a system is becoming congested and then act to restore normal system operation. In this paper, we describe an approach to identify a set of database monitor metrics that can serve as indicators for potential congestion in a specific scenario. We present experiments to illustrate two cases of congestion in a DB2® DBMS and use our approach to derive the indicators.

An approach to managing the execution of large SQL queries

We present an approach to managing the execution of large queries that involves the decomposition of the queries into an equivalent set of smaller queries. The smaller queries are then scheduled such that the work is accomplished with less impact on other, possibly more important queries.

Lightweight problem determination in DBMSs using data stream analysis techniques

Problem determination in a database management system can be a difficult task given the complexity of the system and the large amount of data that must be collected and analyzed. Monitoring the system for this data incurs overhead and has a detrimental effect on application performance. As an alternative to the standard practice of storing the performance data and performing offline analysis, we examine an approach where monitoring data is produced as a continuous data stream and data stream mining techniques are applied. We implement this approach as a prototype system called Tempo on IBM DB2®. Tempo implements Top-K analysis, which is a common task performed by database administrators for problem determination. Top-K analysis typically identifies the set of most frequently occurring events, or the highest consumers of system resources. Our experimental evaluation indicates that Tempo is time and space efficient, incurs low overhead, and produces accurate results.