As we enter the era of big data, a substantial amount of Input/Output (I/O) requests to storage devices are generated, making the maintenance of I/O performance important. Furthermore, I/O performance directly affects the overall user experience in edge devices. However, with the increasing complexity of systems, numerous factors influencing I/O performance have emerged, making it challenging to analyze and explore the overall I/O processing workflow. To address this issue, we introduce IoLens, a visual analytics tool that helps users explore system I/O performance from kernel I/O stack up to virtual file system and storage device drivers. Our tool helps users analyze I/O performance by identifying the overall workload of I/O requests and intuitively identifying anomalies. The effectiveness and applicability of IoLens have been validated through a usage scenario following a system engineer working on system kernels. A user study with four domain experts is conducted to further validate the usability of the tool.

The rapid development of deep learning algorithms, as well as innovative hardware advancements, encourages multi-DNN workloads such as augmented reality applications. However, existing mobile inference frameworks like TensorFlow Lite and MNN fail to efficiently utilize heterogeneous processors available on mobile platforms, because they focus on running a single DNN on a specific processor. As mobile processors are too resource-limited to deliver reasonable performance for such workloads by their own, it is challenging to serve multi-DNN workloads with existing frameworks.