Home Mortgage Company

Market Segment: Online Retail

Introduction

The Company in this case study is the largest online retail home mortgage lending firm in the US, according to National Mortgage News. The company is comprised of several divisions including a loan division, a title company and a settlement services provider. Founded in 1985 the home mortgage company employs more than 3,500 people.

Problem/Opportunity

The Company had approximately 302 cubes in their data warehouse that were being updated every 30 minutes and the database was approximately 250 GB in size. The Company had several Analysis Services (AS) instances. The main instance was on a 64-bit server, while the other two were on 32-bit servers. These instances were separated due to minor differences in data such as data visibility on 64-bit AS cubes, data refreshment and whether or not the data was needed for the daily operational reports. The system was also experiencing frequent crashes of Analysis Services at the time of this engagement. The goal of the Architectural Design Review project completed by Scalability Experts (SE) was to provide recommendations for two key issues:

Develop scalable and highly available server architecture to handle a projected increase in users of 3 – 4 times over the current number of users.
Additionally, the recommended architecture should resolve the current Analysis Services instability issues.

Solution/Capabilities

The main bottleneck was found in the sheer number of cubes, many of them duplicated due to the business requirements. SE analyzed the number of distinct counts that could be reduced and the number of duplicated cubes that could be eliminated by having multiple dimensions for role playing dimensions. One of the recommendations was to consolidate the AS instance on 64-bit with the 32-bit instance server. SE also recommended separating the SQL Server and Analysis Services on their own cluster environments and implement Snapshot Isolation to allow the operational cubes to be refreshed from the snapshot. The resulting architecture and re-design of systems not only eliminated the frequent crashes, but overall system performance and availability was improved significantly.

< Return to Case Studies