With the emergence of the Internet of Things
(IoT) vision, RFID technology has attracted more and more
interests from many industries. In the EPCglobal standard
which standardizes RFID architectural framework and interfaces,
the Electronic Product Code Information System (EPCIS) -
the repository in the RFID network - plays a pivotal role
in storing dynamic and historical RFID trajectory data, welltailored
to supply chain applications. However, in the IoT era,
EPCIS should be capable of storing a large amount of realtime
and continuous time-series RFID data in a scalable manner.
Therefore, in this paper, we analyze the requirements and current
design of EPCIS and report current limitations of SQL-based
approaches for IoT applications. To cope with these limitations,
we explore a NoSQL-based approach for EPCIS in the IoT and
propose a column-family-based data model and customization for
Cassandra-based EPCIS. Using real-world implementation, we
also conduct a performance comparison between MySQL and
Cassandra in different deployment scenarios (i.e., physical and
virtual machines). As a result, we show the benefits of using
Cassandra datastore for EPCIS in the IoT in terms of response
time, throughput, and flexibility.