Columns

UPS-developed messaging software keeps global air hub running smoothly

• By Nick Gray
• June 30, 2003

Build or buy? The question is still asked these days in development. That is even true for middleware, which has been radically altered with the advent of Java and Windows app servers. Middleware was once homebrewed or just-not-brewed and, while it has come to resemble an off-the-shelf item -- broad brush statement, that -- there are still places where building your own publish-and-subscribe middleware system may make the best sense. That was the conclusion of teams at the famous “brown” delivery giant, UPS.

Here, Nick Gray, who directs the infrastructure planning for the distributed computing environment and establishes application architecture best practices and direction for UPS worldwide, takes a look at the value of middleware development within a major company. — The Editors

At UPS, information is critical -- from the size of a package and where it is going, to how it is going to get there. Communication among people and information systems is essential to keep the flow of goods, information and funds moving for UPS customers. This is especially true at Worldport, the company’s global air hub in Louisville, Ky.

Worldport is a 4 million square foot, fully automated facility that is the company’s largest capital investment to date (at more than $1.2 billion) and a linchpin in the global supply chain. More than 2 million express packages from around the globe pass through Worldport every day. It has to run like clockwork. When UPS Information Services initially evaluated the design of the new hub, which called for the consolidation and automation of manual package-sorting functions, the group saw 50 independent applications running on five different operating systems that would have to interact. With the new hub sorting as many as 150 packages every second, and each package requiring multiple interactions among these systems, a middleware system that could support more than 3,000 messages each second was required. Unfortunately, no such off-the-shelf product existed. So UPS turned to its internal software engineers to develop UPS’ Common Infrastructure Services (CIS) and Common Message Environment (CME).

Deciding to take the development in-house

Based on a publish-and-subscribe paradigm, CIS/CME serves as a message broker, tying together the air hub’s multiple systems and forming the backbone of the information systems running in the air hub.

Currently, CIS/CME can sustain up to 6,000 messages per second -- double the capacity Worldport requires -- giving the system room for future growth. Every major system in the hub is connected to the messaging system, and the overall health of the information systems is monitored by the messaging system itself.

To understand exactly how and why CIS/CME was developed, however, it is necessary to understand the fundamental requirement at UPS for all technology to support the core business initiatives. Any technology that is implemented must make UPS operational processes not only cost-effective and efficient, but also customer-friendly. In other words, if the technology cannot ultimately benefit the company’s 7.9 million daily customers worldwide, it is not worth implementing.

Such was the case with CIS/CME. UPS had built and used middleware tools of this type for years with great success. But with Worldport, UPS realized that it needed a faster, fault-tolerant middleware engine and interface. The question was: Could UPS design and implement a system that would meet technology requirements as well as the company’s business goals? UPS IT managers believed they could.

Coordinating the project

UPS decided to pursue development of Messaging Oriented Middleware (MOM), a specific class of software that operates on the principles of message queuing and/or message passing.

MOM permits the reliable transfer of information between dissimilar networks and applications running on those networks; guarantees delivery of messages; ensures that messages are delivered in the order they are sent; and provides secure messaging capability, including access control, message encryption and message privacy. MOM’s benefits include:

* asynchronous messaging for process-to-process and app interoperability,

* systems management capabilities, and

* real-time automation and distributed transaction processing.

The factors influencing the growth of MOM today include the need to integrate legacy systems with e-business systems, the need to extend intra-company processes outside of the company, and the need to respond to rapid business changes -- all factors that were facing UPS as it looked for middleware to meet Worldport’s needs.

Worldport is home to more than 100 miles of computer-controlled, high-speed conveyers; 167 camera tunnels; and enough fiber-optic cable -- 4,482 miles -- to stretch from Louisville to Seattle and back. Hundreds of planes fly in and out of the hub each night, transporting millions of packages, and information about each package is transmitted instantly. Worldport computer systems conduct more than 59 million database transactions per hour. All of this activity needs to be monitored by the middleware system.

Developing the middleware

As UPS began to assemble plans for new middleware to meet these needs, the company realized that for the system to work, and for multiple UPS applications to communicate successfully, it first had to ensure consistency of all message formats, which took buy-in from leaders throughout the company. To accomplish this goal, a group specifically responsible for the development of common message specifications was formed to ensure that the message formats in use were consistent with the company’s needs and well known to all the systems that needed to make use of them.

Before CIS/CME was developed, cross-system, high-volume messaging was left up to each UPS application team to design and build. So when it came time to develop the middleware for Worldport, the team knew it had to approach the problem differently.

Not only would UPS be connecting large numbers of applications with a single middleware product, it would be connecting applications built to run on different hardware, software and operating systems. Because UPS uses a variety of operating systems, it would be a major, and essential, breakthrough to bring these systems all together.

CME was developed as the core message engine; CIS was built around CME, providing a layer between the messaging engine and the applications. CIS was also designed so that it could be switched out for another core middleware engine, such as IBM’s MQ, without impacting its applications. Without this important step, changes to the middleware engine would result in substantial duplicate code changes.

CIS/CME was also made backward-compatible with UPS’ older messaging products, so upgrades of such older products to the new middleware would not require any application changes. This allowed each UPS location to upgrade to CIS/CME on its own schedule, eliminating the need for and the cost of a single, highly coordinated worldwide rollout of multiple applications. Meanwhile, replacing older middleware products with CIS/CME would increase the performance and capacity of the PCs supporting the UPS operations, extending the life of those PCs.

Finally, the team developed CME- View, a customized graphical middleware monitor that is a core part of CIS/CME. By monitoring the status of the many middleware components, CMEView provided the UPS operations staff with a graphical overview of the health of Worldport -- from air and ground traffic, to scanning control systems, to package automation. CMEView also provided complete real-time information on hub activities. This allows the UPS operations staff to quickly recognize problems and correct them before they result in service failures.

By developing CIS/CME in-house, UPS was able to customize the middleware tool, ensuring that it fit the company’s needs. As a result, UPS has total control over the timing of upgrades and the enhancements included in each. And, while the technology is transparent by design, users are noticing a time savings in information processing and smoother operations overall. Users are enjoying a standardized system, one that seamlessly integrates operations throughout Worldport to improve communications and increase efficiency.

Knowing UPS has a flexible tool that does not require a licensing fee is an advantage for the company, as well. It allows IT managers to consider future capabilities, such as expanding the middleware to subsidiaries, without the concern of justifying the cost of a software acquisition.

But perhaps most importantly, CIS/CME more than met its goal of making Worldport run smoothly. UPS could not operate Worldport at the rate it needs to without CIS/CME. The middleware supports moving package destination information from the data center into the hub, handles problems with individual packages during processing, and even helps U.S. Customs agents to identify packages for inspection.

Since UPS developed CIS/CME, other off-the-shelf products have been introduced to the market by several well-known software vendors. But while other companies have developed messaging systems, at the time UPS developed CIS/CME, none could handle the scale of its message requirements.