Seminarium: Systemy Rozproszone
11 grudnia 2008, godzina 12:15,
sala
Many techniques have been proposed to scale web applications. However, the data interdependencies between the database queries and transactions issued by the applications limit their efficiency. We claim that major scalability improvements can be gained by restructuring the web application data into multiple independent data services with exclusive access to their private data store. While this restructuring does not provide performance gains by itself, the implied simplification of each database workload allows a much more efficient use of classical techniques. We will illustrate the data denormalization process based on the TPC-W e-commerce benchmark and show that restructuring its data can provide at least an order of magnitude improvement in the maximum sustainable throughput compared to master-slave database replication, while preserving strong consistency and transactional properties.
We will then turn to resource provisioning for this type of service-oriented applications: given that workloads change over time, how can one continuously determine how many computing resources are necessary to maintain a given level of expected quality of service, and how these resources should be used for optimal utilization? I will show that relatively simple queuing network theory allows one to predict the performance that the service would have if resources were assigned differently, thereby facilitating resource provisioning.
Guillaume Pierre,
Vrije Universiteit