Zhi-Li Zhang's research while affiliated with University of Minnesota Twin Cities and other places

... More recently, with the rapid development of deep learning techniques, the integration of Graph Neural Networks (GNNs) and Recurrent Neural Networks (RNNs) is emphasized to enable accurate spatiotemporal prediction in transportation systems, such as parking occupancy [17], traffic speed [18], flow [19] and also EV charging demand predictions [20], [21], e.g., SGCN [22] combines graph convolutional network (GCN) and gated recurrent unit (GRU) to extract spatial and temporal features of the operating status at EV charging stations, respectively to better assist the prediction. However, the typical convolutional and recurrent structures can be crippled by their inflexibility and ambiguity in weight assignment. ...

... Perhatikan bahwa server cloud yang disediakan oleh perusahaan seperti Amazon dan Alibaba juga didasarkan pada virtualisasi. Namun, dibandingkan dengan VPS, mereka memiliki banyak manfaat lanjutan terkait keandalan, skalabilitas, dan elastisitas [9]. ...

... A novel implementation of Multi-Path Transmission Control Protocol (MPTCP) specifically tailored for SD-WAN environments, referred to as WAN-aware MPTCP, has recently been presented [11]. This implementation aims to optimize the utilization of multiple, heterogeneous WAN transports through aggregation, while also providing fast failure recovery for applications. ...

... Zhang and Zhang [26] Improving network function performance, safety and management Virtualization. An era of all-embracing connectivity has consistently developed, multiple new applications, network protocols, and internet services are supported. ...

... In recent years, cloud computing has been of great interest to the research community [1] [2]. It is used in a variety of applications ranging from healthcare [3] [4] to geosicences [5] and autonomous vehicles [6], [7]. Cloud computing plays a significant role in modern technology as it is actively interacting with other technological fields such as edge [8] and fog computing [9], Internet of Things (IoT) [10], sensor technology [11] and big data [12] [13]. ...

... In many instances, the deployment cannot be optimised, leaving the network redundantly and randomly deployed [1]. Data dissemination is thus quite complicated and difficult in such a network with multiple routes existing between a source and a destination [2,3]. In these networks, data dissemination is usually accomplished through flooding which is very burdensome in terms of bandwidth, energy, and processing [4]. ...

... Additionally, multiple different applications (Chrome, Firefox, wget, curl, etc.) invoke the same application protocol (HTTP). Recent work towards network-application integration [14]- [16] suggests a need to differentiate application processes when performing customization. To allow customization granularity on a per application process level even in cases where targeted processes are yet active, while avoiding transport protocol modification, we have chosen to tap into application messages when they arrive at socket buffers, right before they are passed down to the transport protocol on the sender end, and right after the transport layer finishes processing on the receiver end. ...

... An example use case is the adaptive bitrate (ABR) mechanism [15,36,38] used for online video streaming, where the network performance counters are used to adaptively decide the video playback quality depending on the QoE requirements of the end-users. To enable applications to use such network-related information effectively, various recent works have focused on designing frameworks and APIs for Networkaware Applications (NAA), such as Mobile and Wireless Information Exposure (MoWIE) [42] and Network Exposure Functions (NEF) for 5G networks 1 . In this context, one of the essential aspects that help an auto-adaptive NAA to decide its state is the perceived network throughput of that corresponding application. ...

... At present, the number of user data leakages increased by 15.46% from 2018, and the privacy risk index increased by 26.66% [9]. Users inevitably leave a trail of footprints in the real world while accessing online services from a mobile device [10]. For example, people share various information on Twitter, and even when the original blog is deleted, relevant comments remain on the web [11]. ...

... These applications circumvent OS services (including the network stack), instead they come with their own NIC drivers and perform most of the scheduling and resource management themselves [18,61]. This approach forces the OS to rely on the NIC to implement security (IOMMUs [72,80] or VLANs [75] are not sufficient) and resource-sharing policies [30,44,90]. As a result, the OS becomes just an over-engineered bootloader [27,28,37,87]. ...