The computational sparsity of Mixture-of-Experts (MoE) models enables sub-linear growth in compute cost as model size increases, thus offering a scalable path to training massive neural networks. However, existing implementations suffer from low GPU utilization, significant latency overhead, and a fundamental inability to leverage task locality, primarily due to CPU-managed scheduling, host-initiated communication, and frequent kernel launches. To overcome these limitations, we develop FlashMoE, a fully GPU-resident MoE operator that fuses expert computation and inter-GPU communication into a single persistent GPU kernel. FlashMoE enables fine-grained pipelining of dispatch, compute, and combine phases, eliminating launch overheads and reducing idle gaps. Unlike existing work, FlashMoE obviates bulk-synchronous collectives for one-sided, device-initiated, inter-GPU (R)DMA transfers, thus unlocking payload efficiency, where we eliminate bloated or redundant network payloads in sparsely activated layers. When evaluated on an 8-H100 GPU node with MoE models having up to 128 experts and 16K token sequences, FlashMoE achieves up to 9× higher GPU utilization, 6× lower latency, 5.7× higher throughput, and 4× better overlap efficiency compared to state-of-the-art baselines—despite using FP32 while baselines use FP16. FlashMoE shows that principled GPU kernel-hardware co-design is key to unlocking the performance ceiling of large-scale distributed ML. We provide code at https://github.com/osayamenja/FlashMoE.

When training a deep learning (DL) model, input data are pre-processed on CPUs and transformed into tensors, which are then fed into GPUs for gradient computations of model training. Expensive GPUs must be fully utilized during training to accelerate the training speed. However, intensive CPU operations for input data preprocessing (input pipeline) often lead to CPU bottlenecks; correspondingly, various DL training jobs suffer from GPU under-utilization. We propose FastFlow, a DL training system that automatically mitigates the CPU bottleneck by offloading (scaling out) input pipelines to remote CPUs. FastFlow carefully decides various offloading decisions based on performance metrics specific to applications and allocated resources, while leveraging both local and remote CPUs to prevent the inefficient use of remote resources and minimize the training time. FastFlow's smart offloading policy and mechanisms are seamlessly integrated with TensorFlow for users to enjoy the smart offloading features without modifying the main logic. Our evaluations on our private DL cloud with diverse workloads on various resource environments show that FastFlow improves the training throughput by 1 ~ 4.34X compared to TensorFlow without offloading, by 1 ~ 4.52X compared to TensorFlow with manual CPU offloading (tf.data.service), and by 0.63 ~ 2.06X compared to GPU offloading (DALI).

Apache Spark is a widely used in-memory processing system due to its high performance. For fast data processing, Spark manages in-memory data such as cached or shuffling (aggregate and sorting) data in its own managed memory pools. However, despite its sophisticated memory management scheme, we found that Spark still suffers from out-of-memory (OOM) exceptions and high garbage collection (GC) overheads when wild memory consumers, who are not tracked by Spark and execute external codes, use a large amount of memory. To resolve the problems, we propose PokéMem, which is an enhanced Spark that incorporates wild memory consumers into the managed ones to prevent them from taking up memory spaces excessively in stealth. Our main idea is to open the black-box of unmanaged memory regions in external codes by providing customized data collections. PokéMem enables fine-grained controls of created objects within running tasks, by spilling and reloading the objects of custom data collections based on the memory pressure and access patterns. To further reduce memory pressures, PokéMem exploits pre-built memory estimation models to predict the external code's memory usage and proactively acquires memory before the execution of external code, and also performs JVM heap-usage monitoring to avoid critical memory pressures. With the help of these techniques, our evaluations show that PokéMem outperforms vanilla Spark with at most 3× faster execution with 3.9× smaller GC overheads, and successfully runs workloads without OOM exception that vanilla Spark has failed to run.

Nowadays, deep learning techniques show dramatic performance in computer vision areas, and they even outperform humans on complex tasks such as ImageNet classification. But it turns out a deep learning based model is vulnerable to some small perturbation called an adversarial attack. This is a problem in the view of the safety and security of artificial intelligence, which has recently been studied a lot. These attacks have shown that they can easily fool models of image classification, semantic segmentation, and object detection. We focus on the adversarial attack in semantic segmentation tasks since there is little work in this task. We point out this attack can be protected by denoise autoencoder, which is used for denoising the perturbation and restoring the original images. We build a deep denoise autoencoder model for removing the adversarial perturbation and restoring the clean image. We experiment with various noise distributions and verify the effect of denoise autoencoder against adversarial attack in semantic segmentation task.

As supply chain has become extremely complicated, traditional supply chain management schemes reveal limitations in keeping track of resources for efficient risk management. Recently, blockchain-based solutions for supply chain have shown possibility to overcome those limits. However, numerous practical problems in this research domain are not fully explored up to date. Thus, we delve into one of the most significant problems among them: verification of business logic in transactions. In this paper, we propose a novel distributed ledger system for supply chain, enabled by anomaly detection framework that verifies semantic correctness of transactions based on business context data. Our blockchain data model is tailored to accurately represent events occurred in supply chain based on real-world business standards. In order to facilitate more efficient tracking of provenance, we leverage graph data model to represent supply chain network. On top of the data model, we present smart contract-based anomaly detection framework that verifies whether a transaction is anomalous. Generic rule-based and graph-based detection methods are devised. The feasibility of our proposed model is shown by the system implemented using Hyperledger Sawtooth. We show how our anomaly detection layer can be plugged into the system: how it interacts with other system components, and how overall system flow works with this new function. We evaluate our system with scenario-based simulations. For a number of use cases we synthesized, the correctness and effectiveness of our system are demonstrated.

In recent years, the connected automotive industry has grown explosively. Various connected car services are emerging, such as remote vehicle diagnostics, driver's health monitoring, infotainment, and vehicle safety management. As a result, the number and type of vehicle data are increasing tremendously day by day. However, existing connected automotive solutions have a limitation in that each company manages its own closed data silos. This restricts connected car services from using data sources in various domains. Hence, we propose the GS1 Connected Car, an integrated vehicle information platform, and its ecosystem. We suggest GS1-based automotive data standards for not only in-vehicle data but also all the automotive-related data generated during the lifecycle of vehicles. We provide standardized data collection to EPCIS, the discovery of global automotive services using ONS, IoT Mash-up service between an in-car dashboard platform and IoT devices, video infotainment called GS1 video, and automotive lifecycle management application. We have implemented our platform in a real car by developing an Android-based vehicle dashboard, including service discovery, Mash-up services, and GS1 Video. Also, a mobile application for lifecycle management and Amazon skills for collecting driver's information are developed. Our demonstration and case study show the feasibility of the proposed platform, widening the scope of future connected car services.