Adaptive Mechanisms for Managing the High Performance Web-based Applications
HPCASIA '05 Proceedings of the Eighth International Conference on High-Performance Computing in Asia-Pacific Region
Adaptive self-tuning memory in DB2
VLDB '06 Proceedings of the 32nd international conference on Very large data bases
Physical Database Design: the database professional's guide to exploiting indexes, views, storage, and more
Adaptive Resource Allocation Control for Fair QoS Management
IEEE Transactions on Computers
Adaptive Fair Sharing Control in Real-Time Systems Using Nonlinear Elastic Task Models
IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
A Vision for Next Generation Query Processors and an Associated Research Agenda
Globe '09 Proceedings of the 2nd International Conference on Data Management in Grid and Peer-to-Peer Systems
ACM SIGOPS Operating Systems Review
Efficient load balancing in partitioned queries under random perturbations
ACM Transactions on Autonomous and Adaptive Systems (TAAS) - Special section on formal methods in pervasive computing, pervasive adaptation, and self-adaptive systems: Models and algorithms
| Hi-index | 0.00 |
Load balancing is widely used in computing systems as away to optimize performance by reducing bottleneck utilizations,such as adjusting the size of buffer pools to balanceresource demands in a database management system. Loadbalancing is generally approached as a constrained optimizationproblem in which only the benefits of load balancingare considered. However, the costs of control are importantas well. Herein, we study the value of including in controllerdesign the trade-off between the cost of transient imbalancesin resource utilizations and the cost of changingresource allocations. An example of the latter are actionssuch as resizing buffer pools that can reduce throughputs.This is because requests for data in pools whose memory isreduced immediately have longer access times whereas requestsfor data in pools whose memory is increased must fillthis memory with data from disk before accessed times arereduced. We frame our study of control costs in terms of thewidely used linear quadratic regulator (LQR). We develop acost model that allows us to specify the LQR Q and R matricesbased on the impact on system performance of changingresource allocations and transient load imbalances. Ourstudies of a DB2 Universal Database Server using benchmarksfor online transaction processing and decision supportworkloads show that incorporating our cost model intothe MIMO LQR controller results in a 14% improvement inperformance beyond that achieved by dynamically allocatingthe size of buffers without properly considering the costof control.