Building a Fast Universal Data Access Platform

Chapter 3. Meeting the Performance SLAs for Making Business-Critical Decisions

Pulling data from multiple sources is necessary to make effective decisions about your business; however, these data sources are not as useful if they cannot be queried and cannot return data quickly—data delays result in missed opportunities and failure to identify issues. So companies look for ways to increase the variety and volume of data being used, as well as ways to improve the performance of that data and return insights in speeds that are closer to real time. Let’s look at some of the common methods for improving database performance.

Dynamic Clustering

The first method for improving database performance is dynamic clustering. In short, a central server acts as a delegator, shifting the system’s demands across a bank of worker servers. This distributes the query requests and data returns across multiple server nodes. Each server in the cluster provides data on system performance. This allows the workload manager to determine where to assign requests dynamically as resources become available.

By distributing the query across multiple nodes, the system distributes the workload across multiple servers, thus increasing the processing power and memory devoted to query returns. Because the clustering is dynamic, the delegating server can shift the work among multiple nodes, ensuring that the queries processed are run as efficiently as possible.

Dynamic clustering provides an additional benefit: ...