ParAccel Analytic Database

Published
  • March 15 2010, 5:39pm EDT
More in

REVIEWER: Atul Gandhi, Senior Director, TSG, of Merkle.

BACKGROUND: Merkle is a leading database marketing agency that provides clients with information-based marketing solutions. The $220 million company manages more than 75 marketing databases on behalf of clients and its more than 100 statisticians develop over 800 models annually. Each year, Merkle manages several billion direct mail pieces and informs more than 5 billion marketing decisions.

PLATFORMS: Merkle tapped the ParAccel Analytic Database (PADB) for the newest generation of its KnowledgeLink application, which is a combination of technology, processes and services needed to create and maintain an accurate and comprehensive representation of a customer across multiple channels, business lines and enterprises. In the PADB solution, Merkle leverages a 3-node test environment and a 5-node (B2B) and 8-node (B2C) production environment. For data extraction, transformation and loading, Merkle utilizes Microsoft SQL Server for metadata management and script generation whereas PADB is utilized for large volume data loading and complex transformations.

PROBLEM SOLVED: Merkle set out to improve the performance and accuracy of its KnowledgeLink application while holding firm or reducing costs related to technology and staffing. The application relied upon multiple I/O-intensive processes to integrate data from many different sources, including a major credit verification agency, and performed a myriad of data quality operations before loading the information into various data stores. The prior system had utilized a flat file system developed in-house that ultimately produced significant amounts of redundant data. This required 120TB of storage capacity. Merkle was also seeking to speed the query process and ultimately get information to the downstream applications used by analysts and statisticians, enabling them to develop models faster. The additional time was integral to increasing model accuracy as well.

PRODUCT FUNCTIONALITY: Merkle has migrated to ParAccel the databases it manages for two of its largest clients. On a daily/weekly/quarterly basis, the solution with PADB processes an average of 10MM/100MM/1000MM records, respectively, as well as the reference base accounts for the records, which have been processed with varying information (i.e., records don’t match on one or many components of the data set) over a period of time. The unique set of individuals, households, businesses and addresses are managed and maintained in the PADB once the records go through Merkle’s matching processes and complex algorithms to link and match closely against the existing records in the reference base. Merkle has realized both hard and soft benefits. Performance increased by a factor of nine, and Merkle now enjoys the ability to perform three times the workload in one-third of the time. In addition, the company enjoys a 40 percent savings in price performance. Merkle has recorded a 7 percent increase in accuracy and matching, resulting in more accurate mail files and more targeted and effective campaigns. The soft benefits include more effective data management and a common security framework as well as optimal use of resources and data assets, far easier access to data, and standard and reusable processes.

STRENGTHS: The greatest strengths of the PADB are in its performance, scalability and simplicity. With extremely effective compression algorithms, the overall data load was reduced by 50 percent. The platform has a very low learning curve and sets up quickly. The company provides superior support both in the setup process and as needed; however, Merkle has found ParAccel requires very little ongoing administration. Its open architecture enabled the easy integration with third-party tools and represents an ideal platform for BI workloads.

WEAKNESSES: ParAccel lacks the monitoring tools required for better insight on resource usage models, and if there’s any, it has to be done using command line and by going against many system tables. Workload management is at it infancy state, and while online-reorganization of clusters is on the roadmap, their absence resulted in greater upfront planning and higher initial costs.

SELECTION CRITERIA: Merkle identified columnar data warehouse solutions as providing the optimal performance for the kind of process required by the KnowledgeLink application and considered three columnar database providers in the selection process. The PADB outperformed the other vendors, excelling on price performance. The team quickly identified inherent benefits of the technology that have been proven out and rated high its extensibility. Merkle’s staff also developed a positive working relationship with the ParAccel team.

DELIVERABLES: We are generating complex reports as flat files and sharing with downstream data warehouses.

VENDOR SUPPORT: ParAccel has been very helpful during the product development and the team considers the relationship with ParAccel as a partner rather than a vendor.

DOCUMENTATION: The documentation of the product has been helpful to the team in gaining a better understanding of new and existing features. Even the release notes are very easy to comprehend.

ParAccel Analytic Database

ParAccel

9920 Pacific Heights Blvd.

Suite 450

San Diego, CA 92121

(858) 526-1573

www.paraccel.com

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access