Microsoft SQL Server Customer Solution Case Study FedEx Gains Better Insight into Profitability Metrics with 64-Bit OLAP Solution Overview Country: United States Industry: Transportation Customer Profile Memphis, Tennessee–based FedEx Express is the world’s largest express transportation company and largest independent operating company of FedEx Corporation, with U.S.$17.5 billion in annual revenues. Business Situation An existing UNIX-based business intelligence solution developed by FedEx did not scale or perform well enough to meet all of the company’s analysis needs. Solution FedEx migrated its business solution to Microsoft® SQL Server™ 2000 Analysis Services. When new requirements dictated more scalability, the company to a 64-bit Microsoft Windows® solution. Benefits More granular and easily accessible data Tenfold improvement in query performance 80 percent lower total cost of ownership Six-month return on investment Rapid time to market “SQL Server 2000 Analysis Services running on 64-bit Windows is the only OLAP solution that that we are aware of that can meet our performance and scalability needs.” Scott Mathews, Finance Manager, FedEx FedEx Express (FedEx), the world’s largest express transportation company, needed deeper, more granular insight into factors that influence the profitability of an aspect of its business. The move from a UNIX-based analysis solution to Microsoft® SQL Server™ 2000 Analysis Services delivered a tenfold increase in performance and an 80 percent decrease in the cost of ownership. A subsequent upgrade to a 64-bit Microsoft Windows® operating system—based solution running on an Itanium II–based server delivered an additional threefold performance increase and an 85 to 90 percent decrease in data load times. Situation FedEx Express (FedEx) is the world’s largest express transportation company, providing fast and reliable delivery to 215 countries, including every address in the United States. The company’s global air-and-ground network speeds the delivery of time-sensitive shipments—usually in one to two business days— with the delivery time guaranteed. FedEx manages 882 operating facilities and 10 air express hubs, maintains a fleet of 649 aircraft and more than 42,000 motorized vehicles, and has more than 137,000 employees worldwide. FedEx uses business intelligence solutions to make the data in its information systems available for timely and accurate analysis. Groups across the business rely on those analysis solutions to help them make a broad range of business decisions, including: SQL Server Fast Facts Relational Database Total database size 700 GB Number of tables 1 Size of largest table 700 GB Number of rows in largest table 1 billion Analysis Services Cube size 150 GB Number of OLAP cubes 1 Number of dimensions 16 Number of measures 30 Size of largest dimension More than 1 million members Dimensions with more than 1,000 members 4 Number of users 150 Performance Gains Query response after switch from UNIX to 32-bit Windows (large, noncached query) 900 to 1,000 percent faster Query response after switch from 32-bit Windows to 64bit Windows (large, noncached query) 300 percent faster Cube load times after switch from 32-bit Windows to 64bit Windows 85 to 90 percent decrease Making strategic decisions, such as whether another FedEx Corporation company may be able to better serve a customer. Analysis of marketing campaigns and how they affect shipping volumes and profitability. Determining special services to offer selected customers, such as dedicated account teams or customer service teams. Several years ago, FedEx deployed a commercially available online analytical processing (OLAP) application to help the company assess costs and improve profitability. FedEx originally developed and deployed the business solution on the UNIX platform. The OLAP system ran on a 12-way server and required 24 gigabytes (GB) of RAM. Once per month, FedEx loaded 5 to 8 GBs of new information from its data warehouse into the business solution, which maintained a 25-month moving window of such information. Users queried the business solution using a Web-based OLAP client. Although the business solution worked as intended, FedEx found that the solution could not scale well enough to support all of the company’s analysis needs. The solution began to exhibit data explosion with more than 7 or 8 cube dimensions, resulting in an exponential increase in the disk space used. The result was a rapid falloff in query performance—a behavior that reduced user productivity by requiring users to wait several minutes for the answer to a single question. As a workaround, FedEx was forced to build and maintain two separate OLAP cubes. Because of that separation, the system’s users could not easily integrate those two types of information. In addition to supporting a limited number of dimensions within a single OLAP cube, the “Not only did [SQL Server 2000] Analysis Services enable us to load all data into a single cube, but it also delivered a tenfold increase in query speed.” Bill Berghel, Finance Advisor, FedEx UNIX-based solution could not support a large enough number of members within a single dimension, which limited the granularity of data that could be viewed through the solution. Users could not work at an individual account level, instead having to limit their queries to the customer level, even though a single enterprise customer may have hundreds or thousands of separate accounts. In addition, the solution’s limited scalability had forced FedEx to leave some desirable business metrics out of the OLAP cubes entirely. To meet all of its analysis needs, FedEx needed an OLAP product that could: Scale well past the limits of its existing solution; Provide greater granularity of data; Provide more views into the data; and Support all data within a single cube. That cube would need to support 30 measures and 16 dimensions—three with more than 1,000 members, and a fourth with more than 1 million members. As before, new data would need to be loaded once per month—a task that had to be completed within a weekend to avoid having an impact on user access. And the solution had to allow users to create queries based on any 5 or 6 dimensions and still deliver acceptable query performance. Solution FedEx overcame the scalability and performance limitations of its initial implementation by switching from a UNIXbased OLAP environment to one based on Microsoft® Windows Server System™ integrated server software, including the Microsoft Windows® 2000 Advanced Server operating system and Microsoft SQL Server™ 2000 Analysis Services. The move to Analysis Services enabled FedEx to put all data in a single cube and deliver a tenfold increase in query performance. Some 18 months later, when new business requirements demanded an additional large increase in OLAP performance and scalability, the company again met its needs by migrating from a 32-bit Windows solution to the 64-bit versions of the Microsoft Windows Server™ 2003 operating system and SQL Server 2000 running on Intel Itanium II– based servers. The move to a 64-bit system delivered a threefold increase in query performance over the 32-bit system and an 85 to 90 percent reduction in cube load times, enabling FedEx to rebuild the entire OLAP cube each month. The new business solution is proving useful to a number of groups at FedEx because of the high degree of flexibility it provides in how data can be analyzed—enabled in part by having all data in a single OLAP cube. The solution is used heavily by 150 power users, who are distributed throughout the company’s marketing department, pricing department, and financial planning and analysis group, as well as by users who work for the company’s parent corporation. Data also is imported to other business analysis systems for use by additional personnel. “SQL Server 2000 Analysis Services running on 64-bit Windows is the only OLAP solution that we are aware of that can meet our performance and scalability needs,” says Scott Mathews, Finance Manager at FedEx. “With the level of scalability it provides, it helps determine how we can improve profitability.” Switching to SQL Server 2000 Analysis Services Before switching its solution from a UNIXbased OLAP environment to one based on “Thanks to SQL Server 2000 Analysis Services, we now have a much better, more readily available understanding of how certain metrics affect profitability.” Scott Mathews, Finance Manager, FedEx SQL Server 2000 Analysis Services, the FedEx technology assessment team investigated different solutions, finding that SQL Server was the only option that could meet all of its requirements. The company examined some proprietary analysis solutions, but preferred to go with one of a more open nature so that it could continue to use its existing Web portal. In addition, FedEx saw that a move to SQL Server would significantly reduce the solution’s cost of ownership. FedEx brought in an external technology consultant to assist with the implementation of the new solution. Steps that were required included: 1. Setup of the new hardware (an Intel Xeonbased server with four processors and 4 GB of RAM). 2. Installation of Windows 2000 Advanced Server and SQL Server 2000, including Analysis Services. (The solution was later upgraded to Windows Server 2003.) 3. Connection of the new system to the company’s existing data warehouse—done by using an ODBC driver for the data warehouse environment. 4. Use of the graphical tools provided by SQL Server 2000 to design the OLAP cube and load it with data. 5. Integration of the new OLAP solution with the company’s existing Web portal. Thanks, in part, to the ease of use of Analysis Services, the consultant had the new solution up and running in only three days. “The migration to SQL Server 2000 Analysis Services couldn’t have been simpler,” says Mathews. “We had the new solution running in just a few days.” Cube size remained approximately the same as for the previous solution, even though the new Analysis Solution combined what had been two cubes into one and added a good deal more detail, dimensions, and measures. Analysis Services did not present the same data explosion issues that the UNIX-based solution had exhibited in the past. After Analysis Services was up and running, FedEx spent a month or so tuning the new solution. Although the business solution now contained more information than before and ran on a smaller, less expensive server, it delivered a 900 to 1,000 percent improvement in query performance. Complex queries that took an hour on the UNIX server ran in 5 to 6 minutes, while queries that used to take a minute now ran in 5 to 6 seconds. “People told us that we would never get acceptable query performance from a single cube,” says Bill Berghel, Finance Advisor at FedEx. “Not only did Analysis Services enable us to load all data into a single cube, but it also delivered a tenfold increase in query speed.” Exchange Rate Neutralization Some 18 months after its move to a 32-bit Analysis Services solution, FedEx decided to expand its use of its solution to include the analysis of its international shipments. Not only would this decision require the loading of additional data every month, but also the regular refresh of stored data—as a means of recalculating the cost of past activities using current exchange rates. At the time, FedEx could load a month of data in 6 to 9 hours, meaning that a full refresh of the cube would take approximately one week—an unacceptable amount of time if it were to remain available to users during workdays and still deliver fast query performance. Knowing that the bottleneck was memory, and that the server already “Our move to Analysis Services paid for itself within six months through lower maintenance and operating costs.” Scott Mathews, Finance Manager, FedEx contained the maximum amount of RAM that Analysis Services could utilize under a 32-bit architecture, FedEx decided to explore the migration of its solution to the 64-bit Windows platform. To validate that a 64-bit Analysis Services solution could meet its needs, FedEx implemented a brief proof of concept. Using the 64-bit versions of SQL Server 2000 and Windows Server 2003, Enterprise Edition, which ran on an HP server with four Intel Itanium II processors and 32 GB of RAM, the company proved that it was possible to refresh a month of data in a single hour. In addition, the company tested the number of members it could put into a single dimension, finding that it could load 3 million customers without issue. Microsoft Services assisted with the proof of concept. Microsoft Services Provides Critical Support To help complete the migration to the 64-bit Windows platform, FedEx again enlisted the aid of Microsoft Services. During the sevenweek engagement, a Microsoft Services consultant laid out a solution that first imported the data from the company’s data warehouse into a SQL Server relational database, and then used that database to load the OLAP cube. He also automated the cube loading process, tuning it to take full advantage of all system resources. In addition, the consultant helped the company think through its long-term strategy for data analysis on the Windows platform, sharing knowledge he thought might be helpful with FedEx employees. Today, the business solution resides in the FedEx data center, alongside other missioncritical business solutions. It runs on an HP rx4640 server configured with four 1.5-GHz Intel Itanium II processors and 32 GB of RAM, which is connected through a fiber-channel adapter to an EMC CLARiiON storage area network. Every month, FedEx moves 25 to 30 GB of new data from its 1.4-terabyte data warehouse into a single table on the 64-bit server. At any time, that table contains about 700 GB of data in 1 billion rows. Data in the relational database table is then used to completely refresh the history in the OLAP cube, which maintains a size of approximately 150 GB. A full refresh takes 48 hours—longer than during the proof of concept because the amount of data that FedEx is loading has continued to grow. After the cube is refreshed, FedEx moves it to the old 32-bit Analysis Services environment—since upgraded from Windows 2000 Advanced Server to Windows Server 2003, Enterprise Edition—for users to access. Although users still run their queries against the same server, they are seeing better performance. FedEx has found that the additional memory available on the 64-bit system allows Analysis Services to build a more highly optimized cube, with query times that used to take 10 minutes now reduced to 3 minutes. Benefits Migration from a UNIX-based analysis solution to one based on Windows Server 2003 and SQL Server 2000 gave FedEx the scalability and performance required to meet all of its data analysis needs. The company’s move to a 32-bit Analysis Services environment provided users with more flexibility in how they could analyze and correlate data, along with more metrics and better query performance. And when new business needs demanded even greater scalability, the company’s move to a 64-bit Analysis Services solution provided the performance gains required to apply the same level of granular analysis to the company’s international business. More Granular and Easily Accessible Data With its new solution, FedEx can effectively analyze cost information that is helping the company to make better, faster decisions on pricing contracts, marketing campaigns, what special services to offer customers, and which FedEx business unit can best serve a customer. Because of SQL Server 2000 Analysis Services, users can analyze the forces that affect that profitability in greater detail and from more perspectives, viewing the data in a number of ways to better understand how costs are affected by customers, products, geographies, and other factors. “Thanks to SQL Server 2000 Analysis Services, we now have a much better, more readily available understanding of how certain metrics affect profitability,” says Mathews. “Users can do dynamic, ad-hoc analysis of that data through a readily accessible Web portal, with enough dimensions, metrics, and performance at their disposal to ask new questions or explore new ideas on the fly.” Additional Benefits In addition to providing FedEx with broader and deeper business intelligence, the company’s migration to a Windows-based analysis solution is delivering several other benefits, some of which include: Value. Even though Analysis Services delivers far better scalability and query performance, its total cost of ownership is only 20 percent of what FedEx was paying for its previous UNIX-based solution. “Our move to Analysis Services paid for itself within six months through lower maintenance and operating costs,” says Mathews. Increased user productivity. Improvements in query performance have made the 150 power users of the Analysis Services solution far more productive, allowing them to spend more time asking new questions and less time waiting for answers. Reuse of existing IT assets. By supporting broad standards such as multidimensional expressions, Analysis Services enabled FedEx to use its existing Web portal instead of having to implement a new solution and train users on a new tool set. Rapid time-to-market. The ease of use provided by Analysis Services allowed FedEx to migrate to the Windows platform in just a few days, with an additional 3 to 4 weeks spent tuning the solution and taking advantage of its new capabilities. The migration from a 32-bit system to a 64-bit one took about a week—again with FedEx spending an additional few weeks to take advantage of the solution’s new capabilities. Ease of management. The new business solution is managed by a single database administrator, who can access the system remotely using the Terminal Server feature of Windows Server 2003. For More Information Microsoft Windows Server System For more information about Microsoft products and services, call the Microsoft Sales Information Center at (800) 4269400. In Canada, call the Microsoft Canada Information Centre at (877) 5682495. Customers who are deaf or hard-ofhearing can reach Microsoft text telephone (TTY/TDD) services at (800) 892-5234 in the United States or (905) 568-9641 in Canada. Outside the 50 United States and Canada, please contact your local Microsoft subsidiary. To access information using the World Wide Web, go to: www.microsoft.com Microsoft Windows Server System integrated server infrastructure software is designed to support end-to-end solutions built on Windows Server 2003. It creates an infrastructure based on integrated innovation, Microsoft’s holistic approach to building products and solutions that are intrinsically designed to work together and interact seamlessly with other data and applications across your IT environment. This helps you reduce the costs of ongoing operations, deliver a more secure and reliable IT infrastructure, and drive valuable new capabilities for the future growth of your business. For more information about FedEx Express products and services, visit the Web site at: www.fedex.com For more information about Windows Server System, go to: www.microsoft.com/windowsserversystem Software and Services Hardware Microsoft Windows Server System − Microsoft Windows Server 2003, Enterprise Edition (32-bit and 64-bit versions) − Microsoft SQL Server 2000, Enterprise Edition (32-bit and 64-bit versions) Services − Microsoft Services Technologies − SQL Server 2000 Analysis Services © 2005 Microsoft Corporation. All rights reserved. This case study is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Microsoft, Windows, the Windows logo, Windows Server, and Windows Server System are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. All other trademarks are property of their respective owners. Document published May 2005 HP rx4640 server with four Itanium II 1.5GHz processors and 32 GB of RAM (handles cube loading) Dell PowerEdge 6600 server with 8 Intel Xeon processors and 4.5 GB of RAM (supports user queries)