Union Pacific Railroad ditches its mainframe for SOA
At the heart of Union Pacific Corp.'s railroad operations is an IBM mainframe-based transportation control system that's been chugging along like a hardworking locomotive for nearly 40 years. According to industry experts, it was a pioneering system when first introduced, and it made the Omaha-based transportation giant one of the first companies in the world to make extensive use of online transaction processing technologies.
But while Union Pacific has been able to build on the system's 11 million lines of macro assembler code and make functional modifications to it over the years, the technology has grown obsolete, according to company executives. That has made it tougher for developers to add features in response to customers' business requirements in what has become an increasingly "alert-driven, workflow-based world," says Martin Malley, assistant vice president of information systems at Union Pacific Railroad.
For example, because of slowing sales in the automotive industry, Union Pacific executives wanted the ability to commingle vehicles from different manufacturers on a single railcar, says Malley. The old system wouldn't have been able to handle that change, he says. But thanks to a new distributed network that has been in development since 2006, the company has been able to introduce some of that functionality and much more.
The emerging network, dubbed NetControl, is based on a service-oriented architecture platform that relies heavily on open-source technologies such as the Apache Web server, Hibernate query software and SpringSource's Java application management tools. It is being written in J2EE and runs on Linux. Components of the new transportation control system are being deployed in stages to help Union Pacific better manage more than 90,000 railcars and 32,000 miles of track. The SOA project will cost $150 million to $200 million, and the system is expected to become fully operational by the end of 2013, at which time the company plans to phase out the IBM mainframe.
A System for the Future
The project was conceived in part because Union Pacific was having difficulty recruiting IT workers with macro assembler skills to maintain and enhance the older system, says Senior Vice President and CIO Lynden Tennison. Many IT workers who are familiar with the mainframe-based system are nearing retirement age, and "most colleges and universities don't even teach assembler," he says.
Union Pacific also wanted to develop a system that would improve customers' interactions with the company and help drive internal productivity gains, says Tennison.
In recent years, many freight customers have downsized or outsourced their shipping departments, says Chip Kraft, director of operations planning at Transportation Economics & Management Systems Inc., a Frederick, Md.-based consulting firm. Those companies are now counting on railroads to provide more services that are easier to access.
Union Pacific is working to address those types of requirements in NetControl, which is in the third year of a multiyear design effort and about 30% complete. For instance, more than 99% of the railroad's customer orders are generated electronically over the Web and through EDI feeds, says Linda Brandl, vice president in charge of Union Pacific's 400-person national customer service center. When freight customers sent electronic orders through the mainframe system, information in the orders didn't always match up with data in Union Pacific's records -- for example, there might have been differences between location codes, says Tennison. When that occurred, customer service representatives had to manually edit the orders.
Union Pacific automated many of those processes in the new platform by applying a rules-based system under an electronic shipment services application that was implemented in August 2004. The system helps catch any disparities and "fix them for good," says Brandl. That application has almost completely eliminated the need for customer service agents to make corrections on orders and bills of lading, says Malley, who is the NetControl project leader.
Pressure to Save
Those improvements alone have saved the company several million dollars annually in labor costs, says Tennison. The system has helped Union Pacific reduce its electronic billing failure rate by 59% since it was installed.
It has also enabled Union Pacific's customer service department to work more efficiently, since it no longer has to delay customer orders while manual corrections are being made, says Brandl.
That kind of automation is critical for transportation companies like Union Pacific, because freight traffic sagged during the recession. In its July 23 quarterly financial report, Union Pacific stated that its year-over-year freight revenue had declined 28% during the quarter that ended June 30.
Such pressures are driving transportation companies to push IT-business projects aimed at generating "significant" ROI through efficiencies and cost controls, says Jack W. Plunkett, CEO of Plunkett Research Ltd. in Houston. Those types of projects include the deployment of Web-based systems to help transportation carriers communicate more effectively with customers "while cutting down on the need for human operators," Plunkett adds.
Union Pacific didn't conduct an ROI analysis on NetControl, because the project was viewed as a risk mitigation strategy to replace a 40-year-old system that has become difficult to maintain but is nonetheless core to the company's business, says Tennison. Still, he says company executives have identified cost savings generated by NetControl.
Tennison says Union Pacific considered other options besides building a completely new system, including the possibility of using on-demand cloud-based services, but they didn't meet the company's needs. "Although we do see huge swings in capacity, we don't in volume changes. We didn't see the economics, and we don't have the wide swings in compute requirements" to justify a shift to on-demand computing services, he says.
The NetControl project also gives Union Pacific a chance to make its transportation services better for its customers. For instance, whereas the older system offered freight customers the ability to track and trace their orders, the new SOA-based platform will enable customers to see their actual itineraries. That will be made possible by more tightly integrated Internet capabilities and new applications that tie in more directly to customers' business processes, says Brandl. She notes that the 220-person NetControl project team recently began planning the development of that functionality.
Because the NetControl system is being phased in over time, Union Pacific's freight customers already have increased visibility into their orders. For example, customers no longer have to call or fax in their orders a few days ahead of time with the number of freight cars they'll need for a single shipment or set of shipments, says Brandl. In 2005, the NetControl project team developed an application called Total Car Management that made it possible for customers to place orders via the Web and use their browsers to see when freight cars have been applied to their orders, says Brandl. The application has helped Union Pacific reduce the amount of manual car orders by 75% since it was installed, according to company officials.
Union Pacific's shift away from the mainframe system's green-screen interface has also benefited many of the younger employees. For instance, Brandl says it takes 10 weeks to train a customer service rep on the old system, but it's expected to take just six weeks once the new system, which is much more intuitive, is full implemented.
Younger people "have grown up in a texting environment with MySpace and Facebook," says Brandl, so many of the company's younger staffers and customers "expect to be able to go to the Web to find what they're looking for without having to use some special code to get there" as they would with the green-screen system.
Still, the NetControl project team had to make sure that the new GUI wouldn't be a complete shock for those employees who had become accustomed to the look and feel of the old system. To strike a balance, developers have created a set of browser-based style sheets that allow end users to choose their GUI preferences, says Tennison.
It helps that the project team has introduced functionality in "baby steps," says Tennison. From a cultural perspective, he says, "that made it a lot easier, maybe because we didn't go too fast." Instead, the company has laid new track for NetControl while the mainframe is eased into the boneyard.
At a Glance
Company: Union Pacific Corp.
Business: Its principal operating company, Union Pacific Railroad, provides freight transportation services to 23 states in the western two-thirds of the U.S.
Company stats: Revenue of $18 billion in 2008; net income of $2.3 billion in 2008
Project champions: Initially, it was Lynden Tennison, senior vice president and CIO. The current project leader is Martin Malley, assistant vice president of information systems at Union Pacific Railroad.
Size of project team: 220 people, including offshore developers
Size of IT staff: Just over 1,400 people, including 400 in telecommunications operations
Project investment: Union Pacific plans to invest roughly $200 million in a new transportation distribution network that's being completed in stages through 2013.
Project payback: By automating customer billing processes, Union Pacific has dramatically reduced the amount of manual edits needed to process customer bills and is saving millions of dollars in labor costs.
Hoffman is a freelance writer in New York. Contact him at firstname.lastname@example.org.
When CIO Lynden Tennison and other members of the Union Pacific IT team began exploring technical platforms for hosting the NetControl system, they identified a few must-haves: The system would have to be long-lasting, able to scale effectively and able to support a "five-nines" (99.999% uptime) environment.
"We designed an engineering model of a loosely coupled server environment, much like you'd find at a Google but not quite as sophisticated," says Tennison of the initial design efforts, which began in early 2003.
The team spent about two years running test evaluations and mock-up models. "Once we understood enough about the technical framework, we started in earnest on the redesign of the core business in mid- to late 2006," Tennison adds.
Project team members opted for a distributed network because they wanted a system that could efficiently handle compute-intensive applications, such as those that build optimization models for locomotives and railcars, says Tennison. "We wanted to be able to throw a lot of horsepower at these problems and not be constrained by a big, monolithic engine," he says. The team ultimately decided to run the system on Linux.
For Martin Malley and members of the 220-person NetControl project team, one of the biggest challenges has been determining which features of the old system have to be preserved -- and do that without the aid of documentation on how various business processes in that system have evolved, says Malley.
To do that, the team is applying an agile model where applications are designed three to six months ahead of the software development schedule, says Malley. Application delivery teams work in one-month shifts, and new releases are generated each quarter.
That approach tends to work well for service-oriented architecture initiatives, says Randy Heffner, an analyst at Forrester Research Inc. With an SOA-type framework, says Heffner, "there are opportunities to automate things and rethink how certain processes work."