CASE STUDY - Alvion Scales Data Analytics and Distribution
- By Bojan Belovic
- October 18, 2007
Commentary by Bojan Belovic, Database Administrator, Alvion Technologies
Alvion Technologies, Inc. provides outsourced data management services to some of the largest marketing data list owners in the world—names such as Axiom, Equifax, Experian, and Dun & Bradstreet. Alvion sits squarely in the crossroads of e-commerce and business intelligence, leveraging their exclusive technologies, extensive very large data warehousing (VLDW) expertise, and marketing know-how to provide customers with fast, reliable marketing data.
Business intelligence is the lifeblood of Alvion and companies like it. The data they collect, often from multiple sources, must be aggregated and analyzed if it is to provide value. That requires data warehousing and analytics solutions that ensure:
- The capacity to handle increased query demand, especially for ad hoc queries
- Access to hundreds or thousands of simultaneous users
- Reduced data latency, sometimes providing nearly real-time updating
- The ability to grow with increasing data volumes
- Lower hardware, administration and maintenance costs
- No downtime
- High levels of security
Alvion’s customers’ datasets range up to 190 million records and 200 attributes. Combined, Alvion manages about a terabyte of data for customers as an outsourcing service. Individual customers submit anywhere from 5,000 specialized records of information, to 120 gigabytes of data. Alvion then runs customer-specific data transformations and uploads it to their production servers for access by end users—widely distributed customers of the original data owners. Data update cycles vary from daily to annually.
End users of this data typically run queries and counts to determine optimal selection criteria for specific direct marketing programs. This means that a typical user session includes numerous queries, run with slight variations, to find the strongest candidates for final data selection. As the number of queries a user makes increases, so does the need for speed.
“We use IQ in production because its biggest strength as far as we are concerned are those really, really fast counts,” said Bojan Belovic, database administrator for the Alvion system. “The response time is absolutely the number one reason we are using Sybase IQ. We are talking about an order of magnitude difference.”
Alvion’s success in recent years has translated into significant growth:
- More data hosted from new customers
- Growth in existing customers’ record counts
- Increasing user sophistication and query complexity, and overall data mining goals
- Four-fold end-user growth
Throughout this period, the number of daily ad hoc queries has increased from 2,000 to 7,500—one-third of which are highly complex—while the number of registered users has jumped from 12,000 to 50,000. The system has scaled easily to these new levels, thanks to Sybase IQ, which has also supported the overall tremendous growth in data volume via its unique data compression capability. With Sybase IQ, disk space savings add value, especially considering the cost of the large, high-speed drives required for high-use VLDWs. Plus, the extra drives typically used to tune-up database performance were not needed.
Until now, Alvion divided its data warehouse tasks between queries and order fulfillment. Alvion is looking to consolidate all operations in a single environment, which will mean increasing the size of the production database by roughly 100 percent. Amazingly, Sybase IQ provides such effective compression that the new data will not affect the system’s performance or query speed. However, this change will dramatically improve Alvion’s competitive advantage by shortening data fulfillment times while maintaining the trademark query speeds so appreciated by its customers.
Belovic believes that Sybase IQ is up to that task, too.
“Our fast query times are what defines us in this industry. We are building in more complexity with Sybase IQ as the backbone, while maintaining lightning-fast response times. That’s going to take us a long way."
This article originally appeared in the issue of .