Current Filter: Storage>>>>>Case study>

   

Cartal-Rijsbergen Automotive (CRA) has been active in the Netherlands for many years as a provider of automotive parts, equipment, tyres, tools and accessories supporting garages, mechanics, dealers, specialists, custom fitters, rebuilders, fleet owners, insurance agencies, etc. With headquarters in Amsterdam, 18 sales locations throughout Holland, and two large central warehouses, CRA has earned a competitive advantage of wide product availability and fast delivery that has resulted in high customer satisfaction, brand loyalty and significant return business.

Supporting this after-market business is an IT infrastructure that features a Microsoft SQL Server platform for such critical database applications as data warehousing, data mining and OnLine Transaction Processing (OLTP). In its introductory stage of implementation a few years ago, this environment served the needs of CRA and its customers, but as more products were added to the data warehouse, more data analysis was required to improve business decisions, and more orders were secured, data traffic increased considerably and disk I/O performance became challenged.

In mid-2013, the IT department initiated an infrastructure assessment to evaluate its hard drive-based storage area network (SAN) and find the best way to overcome their SQL Server performance bottlenecks. This included an evaluation of solid-state drives (SSDs) that use NAND flash memory to store data, and with no moving mechanical heads like hard disk drives (HDDs), handle random data access with ease to deliver I/O performance comparable to a large SAN array with hundreds of HDDs implemented.

KEY DATABASES
The SQL Server-based data warehouse provides CRA sales locations and customers with up-to-date online information on product pricing, assortment, stock availability, specifications, features, benefits, warranties, rebates, discounts, etc. The accuracy of this information and the ability to deliver it in real-time can make the difference in securing customer orders and providing a heightened user experience.

In conjunction with the data warehouse, CRA implemented OLTP databases to facilitate and manage all online transactions whether they originate from a CRA sales location or directly from the customer. These databases capture information surrounding a sale and enable the data to be segmented, grouped, stored and retrieved for a specific use-case or analysed as business intelligence (BI).

CRA also implements data mining to sort through SQL Server data and find undiscovered usage patterns and relationships that help them make more informed business decisions. The business intelligence obtained from data mining has enabled CRA in the analysis of sales, products, shipping and routing, sales territories, Enterprise Resource Planning (ERP), forecasting and financial projections, to name a few.

Overall the CRA enterprise includes six key SQL Server databases that support data warehousing, data mining and OLTP applications, with the largest data warehousing database approximately 200GB in size. With millions of database records captured, excessive combinations of product inquiries, online purchases, data analysis and/or real-time data warehouse modifications adversely affected CRA's HDD-based SAN resulting in system strain and non-optimal SQL Server performance requiring an improvement in read/write performance.

INFRASTRUCTURE ISSUES
The Cartal-Rijsbergen Automotive SQL Server system configuration included a Dell PowerEdge R-820 server with RAID 5 capabilities, a Dell NX-3100 networked attached storage (NAS) device and a Dell EqualLogic SAN Array with 8 HDDs included. While the PowerEdge server is capable of handling hundreds of thousands of input/output operations per second (IOPS), each HDD in the SAN Array can only deliver a few hundred IOPS of performance.

As a result, the HDDs cannot keep up with SQL Server demands even if more HDDs were added to CRA's system configuration. As multiple SQL Server workloads run concurrently in the enterprise, strong random access is necessary to accelerate I/O and is the shortcoming of HDDs as they require their mechanical heads to move from location to location, limiting each drive's physical ability to quickly read or write random data. IOPS performance and latency are slowed until data is found and accessed.

Page   1  2