Dec 14, 2009Every time a product equipped with an EPC Gen 2 RFID tag is read by an RFID interrogator, an "event" can be recorded. An event can contain multiple attributes, such as the EPC number, time, location and business process involved. EPC Information Services (EPCIS), a standard developed by EPCglobal, enables the exchange of RFID event data between supply-chain partners, providing visibility into the supply chain. Retailers, for example, can use event data to improve the replenishment process, leading to lower inventory levels and reduced out-of-stock situations.
Today, the RFID data volumes generated and stored in an EPCIS system are relatively low. But that will change as more retailers adopt item-level tagging. To replenish a typical retail store in Europe, for example, roughly 300 pallets containing 20,000 cases, which, in turn, contain 150,000 items, are moved through a supply chain each week. Some estimates claim that one retailer could generate 15 terabytes of data per day.
To determine whether the technology available now is sufficient to handle these massive amounts of data and enable the exchange of information between supply-chain partners, the Auto-ID Lab St. Gallen/ETH Zurich developed a data volume simulation tool; it can estimate the emerging data patterns for different tagging granularities, depending on supply-chain parameters. Researchers and developers can use the tool to estimate expected data-traffic when designing RFID event databases, discovery services (basically, search engines for EPCIS databases) and access-control frameworks.
Our research found that current technology should be able to cope with several terabytes of retailer-generated data. That RFID data volume will be in the range of today's point-of-sale data, which today's network infrastructure can handle. Still, challenges remain.
Because most EPCIS events are represented in XML format, common compression ratios of more than 90 percent are feasible. This means the transmission of data is unlikely to be a problem, especially because few business cases will require full access to all data in real time. But problems could arise if data is shared with multiple partners in a fully synchronized way—that is, continuously as opposed to on demand. A retailer, for example, might need real-time information from its logistics, warehouse and manufacturing partners to make decisions that could minimize processing time and avoid out-of-stocks.
Participants will have to consider the balance between short-term storage and long-term archiving. We believe most of the data residing in EPCIS systems can be purged at a specified time—for instance, after selling the item or when a one-year warranty has expired.
Another critical issue is security. The European Union-funded BRIDGE project is working on how to secure information-sharing systems for EPCIS data, but our study did not consider the data overhead arising from security checks.
Although our simulation tool is based on the characteristics of a typical retail supply chain in Europe, the insights it provides could be valuable to researchers and designers developing EPCIS databases for other industries.
Florian Michahelles is associate director of the Auto-ID Lab St. Gallen/ETH Zurich and a deputy director of information management at ETH Zurich. Alexander Ilic is a senior researcher at the lab.
