PUMA Case Study
PUMA uses the Zing™ JVM for Core System Performance and Scalability
PUMA is one of the world’s leading Sport Lifestyle companies that designs and develops footwear, apparel and accessories.
PUMA starts in Sport and ends in Fashion. Its Sport Performance and Lifestyle labels include categories such as Football, Running, Motorsports, Golf and Sailing. Sport Fashion features collaborations with renowned designer labels such as Alexander McQueen, Mihara Yasuhiro and Sergio Rossi.
The PUMA Group owns the brands PUMA, Cobra Golf and Tretorn. The company, which was founded in 1948, distributes its products in more than 120 countries, employs more than 11,000 people worldwide and has headquarters in Herzogenaurach/Germany, Boston, London and Hong Kong.
PUMA uses a variety of Java-based applications for their web presence, stores and manufacturing operations. The PUMA custom product and supply chain application has about 1 million lines of code and 150 Enterprise Java Beans. During peak times the system reads/writes about 55 million records a day. The company’s rapid growth and innovation as one of the few true multi-category global brands caused them to reach the limits of what their current JVM and Java infrastructure could deliver.
Core Applications Face Bottlenecks
The most important systems experiencing issues were the Product Information Database (PID) and the global Supply Chain Management solution known as the Global Business System (GBS). PID is the source of all article data created and maintained by the design, product management and sourcing groups that is used by ERP, warehousing, the online store and all their digital catalogs. GBS is the core backbone for orders from PUMA subsidiaries. GBS in turn places orders with suppliers, PUMA factories and logistics.
Purchasing agents and other users of these systems wanted to be able to update thousands of orders at a time. However, each user can kick off this process multiple times on top of others’ ‘normal’ usage. This memory-intensive task led to sudden allocations of 15 – 20 GBytes of memory in just a few seconds. This was enough to trigger ‘stop-the-world’ garbage collection pauses in their current JVM that created long order processing delays and frustrated users. And, in batch mode, some operations could never finish.
Azul Technology Resolves Long Processing Time
PUMA turned to Azul Systems for help. They originally implemented the Azul JVM and Azul’s Vega appliances to resolve initial performance and garbage collection issues in their core systems. With Azul, batch order processing completed consistently in only 14 minutes, free of GC pauses. The system was also able to scale linearly, allowing PUMA to increase loads while still enjoying predictable, consistent response times. The system is now much more stable, allowing PUMA to support hundreds of users running hundreds of processes simultaneously.
Bringing Performance Improvements to Other Applications
After eight years of successfully using Azul appliance technology, PUMA evaluated and began implementing the new Zing JVM. Zing implements Azul’s leading-edge Java performance technology in a pure software-only package. Zing is designed and optimized for deployment on Linux and x86-64 servers.
Zing delivers the performance improvements, pauseless operation and scalability PUMA experienced with Azul’s Vega appliances, but allows more flexible deployment options and increases the number of applications that can benefit. Also, Zing fits seamlessly into PUMA’s virtualized infrastructure. In addition to consistent application performance, Zing also provides enterprise-class monitoring and management tools. Zing Vision provides PUMA IT staff with always-on production monitoring that is compatible with their existing monitoring tools. Zing Vision allows IT to find and resolve production issues quickly. Zing’s management tools automatically and dynamically allocate additional resources to application instances based on real-time demand.
Zing is Now the PUMA Standard JVM
PUMA has now standardized on the Zing JVM for all their J2EE headquarters applications. With Zing, PUMA can focus on growing their business and adding innovative new products and experiences for customers without worrying about application performance.
Unpredictable garbage collection pauses caused unacceptable processing times for core applications.
PUMA is migrating their supply chain, product information and other applications to the Zing JVM to provide faster processing and a better user experience.
Predictable response times
Lower capital costs and TCO
Elimination of GC pauses
Simple to implement with no code changes
Increased business flexibility to add innovative web features and expand operations
Instrumented JVM with built-in monitoring
“Azul has been a key part of our solution stack for eight years. In all that time we have not experienced a single issue related to our Java infrastructure or garbage collection.”
– Christian Burger Senior Head IT Custom Solutions, PUMA