Home   People   Publications  
 

Heterogeneous Distributed Shared Memory for Lightweight Internet-of-Things Devices [abstract] (IEEE Xplore, PDF)
Bongjun Kim, Seonyeong Heo, Gyeongmin Lee, Soyeon Park, Hanjun Kim, and Jong Kim
IEEE Micro, Volume 36, Number 6, November 2016.

Despite the promising applicability of the Internet of Things (IoT) such as home automation and health monitoring, developing an IoT application is laborious because programmers should explicitly manipulate communication for shared data across heterogeneous distributed IoT devices. Recent proposals simplify the communication management with graphical views or language extension, but still require programmers to explicitly manage the communication. To liberate programmers from the explicit communication management, this work revisits heterogeneous distributed shared memory (HDSM) for IoT devices, and proposes a compiler-runtime cooperative HDSM framework, called IoT-HDSM. The compiler unifies heterogeneous memory layouts of different IoT devices, and the runtime system provides a shared memory view for the devices. To support lightweight devices that cannot execute the runtime, the compiler automatically inserts explicit communication instructions into the IoT application. More- over, to efficiently control concurrent memory accesses without locks, this work also proposes a new annotation called memory branch that provides a memory snapshot for each device. This work implements 3 IoT services with IoT-HDSM, and shows that IoT-HDSM simplifies shared data management without harming QoS.