LESSON - The Analytics Server: Powering On-Demand Analytics from the Ground Up
- By Phil Bowermaster
- October 18, 2007
By Phil Bowermaster, Worldwide Product Marketing Manager, Sybase IQ
When looking to design and implement a system that will meet their customers’ requirements, companies that provide data analytics services face a number of significant challenges. They need a solution with the capacity to handle heavy and growing query demand, especially for ad hoc queries. They need a solution that can be accessed by hundreds or thousands of simultaneous users without a significant impact on performance. They need to reduce data latency, sometimes providing updates at or near real time. Of course, they also need to minimize downtime and maximize security—all while lowering hardware, administration, and maintenance costs.
These demands aren’t easy to achieve, and those who provide data analytics services often run into roadblocks trying to achieve them using traditional approaches to data warehousing.
For example, a company providing this type of service will deploy a solution built on a standard relational database management system. We’ve seen a number of instances where companies find that, when implementing an on-demand data analytics service on a standard RDBMS, complex queries on large volumes of data run very slowly. This happens because standard relational database management system (RDBMS) solutions are designed for transactional, rather than analytic, environments. They store data in tables by row, where each row is effectively equivalent to a transaction. Querying such a database means grinding through many entire rows of data, when all that the user is looking for is a few select items found in some of the columns.
Another common approach is to use a data warehouse appliance. The typical appliance solution is designed for analytics, so it will offer a significant speed advantage over a transactional database where individual queries are concerned.
Companies using this approach can run into difficulties, too. The fact that appliances run on proprietary hardware limits a company’s options in defining their own environment. They may also find that the appliance approach limits their ability to scale up to support multiple users, and that multiple queries from hundreds or thousands of users can choke the system. To make matters more difficult, database administrators are often limited in the amount of tuning they can do for specialized applications.
Companies that depend on fast, ad hoc query responses by many users accessing large volumes of data need a different kind of solution, one that’s built from the ground up to support on-demand data analytics. This is where an analytics server comes in.
An analytics server, such as Sybase IQ, takes a fundamentally different approach to data analytics. It stores and accesses data in tables by column. This serves to index automatically the entire database, because query selection criteria are defined by column. This greatly reduces the amount of data that needs to be read to respond to a query and dramatically speeds up response times.
Because of these advantages, an analytics server supports extremely fast query speeds without limiting:
- The amount of data available for analytics
- The number of concurrent users and queries
- The launching of new applications
- The complexity of queries
- The number of access windows to data
- Return on investment
In addition to its edge in query response times, an analytics server such as Sybase IQ offers a far better return on investment than other data warehousing and analytics solutions. Because of its speed, it requires fewer hardware resources than other options. Its column-based approach also reduces the need to pre-aggregate data and tune the database, saving significant administration time. All of which goes to explain why, increasingly, we’re seeing the leading players in data analytics services turning to the analytics server as the platform of choice for delivering on-demand analytics solutions.
This article originally appeared in the issue of .