مدیریت حجم کاری در سیستم های مدیریت پایگاه داده: یک طبقه بندی Workload Management in Database Management Systems: A Taxonomy

1 INTRODUCTION database workload is a set of requests that have some common characteristics such as application, source of request, type of query, business priority, and/or performance objectives [72]. For both strategic and financial reasons, some business organizations are consolidating multiple individual database servers onto a shared database server to serve as the single source of corporate data [3] [9]. This trend of database server consolidation means that multiple types of workloads are simultaneously present on a single database server. These workloads may include on-line transaction processing (OLTP), which consists of short and efficient transactions that may require only milliseconds of CPU time and very small amounts of disk I/O to complete, as well as Business Intelligence (BI) workloads [31], which include longer, more complex and resource-intensive queries that can require hours or an even longer time to complete. Workloads submitted by different applications or initiated from distinct business units may have unique performance objectives (goals) that need to be strictly satisfied. The performance objectives of a workload are normally derived from a formal service level agreement (SLA). On a shared database server, there is an interdependence among the concurrently running workloads that results from the workload’s competition for the shared system resources, such as system CPU, main memory, disk I/O, network bandwidth and various queues. If a workload, e.g., an operational BI workload, is allowed to consume a large amount of shared system resources without any control, the concurrently running workloads may have to wait for the workload to complete and to release its used resources, thereby resulting in waiting workloads missing their performance goals and the entire database server suffering degradation in performance. As workload requests present on a database server can fluctuate rapidly among multiple types, it becomes impossible for database administrators (DBAs) to manually adjust the system configurations in order to maintain the workload’s performance requirements during their run time. Thus, workload management becomes necessary and critical to effectively control the processes of different types of workloads and manage shared system resources to achieve a set of per-workload performance goals in a complex workload mix environment. Workload management is the discipline of effectively monitoring, managing and controlling work flow across computing systems [8] [74]. In particular, workload management for database management systems (DBMSs) is the process or act of monitoring and controlling work (or requests) executing on a database system in order to make efficient use of system resources in addition to achieving any performance objectives assigned to that work [3]. Thus, the primary goals of workload management for a DBMS are: 1) to maintain the DBMS running in an optimal state, i.e., neither under-loaded nor overloaded, 2) to ensure that all workloads meet their desired performance objectives (if any), and 3) to balance resource