Providing simply the best, most up-to-date insider data about nearby dealerships, market prices, sales, and car specifications, Edmunds delivers paid advertising for car manufacturers as well as leads- and subscription-based services to dealers
Moving Data Closer to the Business with a NoSQL Solution
CATEGORY
Case Study, Consumer
With Edmunds, potential customers get all of the essential information about the car they are shopping for before they ever step through the door of a dealership.
Edmunds approached EPAM to find a solution for replacing the expensive, hard-to-maintain Oracle Coherence in-memory data grid that had been powering their website for five years and simultaneously limiting the company’s platform evolution. After one-and-a-half years of hard work and collaboration with Edmunds on the dataset migration project, EPAM successfully completed the replacement of Oracle Coherence with MongoDB. The engagement resulted in the lasting business benefits for the client.
Key Challenges
- Replacing the expensive, hard-to-maintain Oracle Coherence in-memory data grid that had been powering Edmund's website and limiting the company's platform evolution
- Migrating the data from one large cluster feature multiple datasets consisting of 4-24 Coherence nodes each
- Delivering an open-source, fast-performing NoSQL that organizes, embeds and monitors its data based on use cases
- Integrating the solution into the existing platform, keeping APIs intact without jeopardizing performance
Solution Highlights
- The first step in finding the proper solution for Edmunds was the launch of a joint investigation into which database would best fit the client’s needs. This resulted in the NoSQL market analysis phase, during which EPAM analyzed roughly 15 different data storages characterized as distributed in-memory data grids, document-oriented databases, and key value databases. EPAM then ranked the storages by criteria such as licensing and operational costs, performance, applicability, reliability, ease of maintenance, and others.After conducting comprehensive research on each storage solution, EPAM chose three candidates for use in proof-of-concept testing. Our team started to run multiple learning loops in order to choose the best candidate to replace Coherence and, based on the results, chose MongoDB as the most balanced solution for the client. In order to meet strict performance requirements throughout the migration, we also applied a combination of Microservices and Direct Data Access Architecture.With the right storage solution selected and the data architectures in place, the EPAM team was ready to begin the migration of Edmunds’ core datasets. We seamlessly migrated the following datasets to MongoDB:
- Vehicle: a configurations, specifications, and pricing dataset
- Region: a DMA, geo regions, and incentive region dataset
- Incentive: a manufacturer incentive and rebate dataset
- TMV: a dataset that powers the True Market Value calculator
- TCO: a dataset that powers the True Cost to Own calculator for new and used cars
- Partner: information about Edmunds partners and contracts
- Rules: rules for Dealer and Inventory data processing
- Maintenance: car maintenance actions, recalls, and service bulletins dataset
- Throughout the migration process, our team worked diligently to ensure that Edmunds’ front-end applications and internal tools would be able to access the data via the shared REST and Java APIs. We also implemented a comparison tool to measure performance of numerous UI pages and REST responses across different environments to make sure that performance was not affected while moving from Coherence to MongoDB.
The Results
After one-and-a-half years of hard work and collaboration with Edmunds on the dataset migration project, EPAM successfully completed the replacement of Oracle Coherence with MongoDB in December 2015. This project is another in EPAM’s very impressive series of collaborations in partnership with Edmunds – a partnership made stronger by EPAM’s ability to accomplish the following:
- Refactor the client’s data structure so deeply that the following improvements became apparent:
- Bulkdata load times improved 3 to 5 times
- Nodes size scaled down from 12 - 24 to 4 - 8
- True Market Value generation time was cut in half from 100 minutes to 50 minutes
- Respond to the client’s needs with professional analysis of the business domain and custom IT architecture to produce the following infrastructure savings:
- 200 fewer hosting companies utilized
- $10,000 to $12,000 per month saved (based on 75% utilization)
- Network bandwidth reduction from 6 Gbps to 3 Gbps for vehicle storage
- Enabled microservices architecture
Technologies Used
- MongoDB
- Oracle Coherence
GET IN TOUCH
Hi! We’d love to hear from you.
Are you ready to design the business models of tomorrow?
