Consensus protocols are essential in distributed systems for fault tolerance, significantly impacting the performance and availability of the upper systems. Recent works have leveraged Remote Direct Memory Access (RDMA) for its superior performance and one-sided primitives to accelerate consensus protocols. Beyond performance, RDMA's processor bypass feature allows servers to handle network requests even when processors fail, enabling a fine-grained failure model where servers remain operational despite processor issues. This fine-grained failure model can help enhance the availability of existing consensus protocols, which traditionally assume processor failures lead to complete server crashes. In this paper, we introduce R2aft (RDMA+Raft), the first RDMA-based consensus protocol fully adapted to this model. Beyond the availability benefits, R2aft also integrates specific optimizations to capitalize on RDMA's high performance. Using the YCSB Workload, we show that R2aft incurs only a 12.2% latency overhead while maintaining throughput comparable to Mu, the state-of-the-art RDMA-based consensus protocol.

Accurate prediction of drug-target interaction (DTI) is crucial for drug discovery. In deep learning (DL)-based approaches, achieving robust and generalizable DTI prediction hinges on informative feature representations and effective modeling strategies. However, existing methods often depend on prior domain knowledge, which may limit their adaptability to unseen samples. Moreover, many approaches fail to exploit the complementary nature of multimodal biological data. To address these limitations, we propose MMF-DTI, a novel multimodal fusion framework that synergizes pretrained large language models (LLMs). These LLMs are employed to extract generalized representations, enhancing the model's capability to generalize to previously unseen molecules. To effectively integrate these modalities, we propose a cross-modal attention module that facilitates interaction between sequence-based and structural features. In comparison with other baselines, MMF-DTI demonstrates a consistent improvement. We further demonstrate the interpretability by aligning attention weights with biophysical interactions, validated on the 1M17 protein.

Drug-target interaction (DTI) prediction is a critical task in drug discovery. However, the differences in size between drugs and proteins present significant challenges in accurately predicting binding sites. Additionally, the issue of modality imbalance, which arises from modality learning biases, undermines the contribution of multimodal representations to DTI prediction. To address these challenges, we propose DDGR-DTI, which is based on an innovative Decoupled Dual-Granularity Framework and a Rebalanced pyramid network (RPN). This framework divides the DTI task into two levels of granularity. The macro level, which decomposes it into subtasks based on modality, and the micro level further decomposes the representation within each subtask. Furthermore, the dual-stream attention module is utilized to perform fine-grained substructure-level interactions within each subtask, thereby enabling accurate identification of binding sites. Simultaneously, we employ an RPN, which effectively alleviates the bias towards the dominant modality in multimodal fusion through a hierarchical aggregation mechanism, emphasizing the synergistic advantages brought by modality balance. Benchmark results demonstrate that DDGR-DTI outperforms existing state-of-the-art models in both prediction performance and generalization ability. Availability: The source code and dataset can be found at https://github.com/ZZUzy/DDGR-DTI.

Point cloud registration, a fundamental task in computer science and artificial intelligence, involves rigidly transforming point clouds from different perspectives into a common coordinate system. Traditional registration methods often lack robustness and fail to achieve the desired level of accuracy. In contrast, deep learning-based registration methods have demonstrated improved accuracy and generalization. However, these methods are hindered by large parameter sizes, complex network architectures, and challenges related to efficiency, robustness, and partial overlaps. In this study, we propose a lightweight deep learning-based registration method that captures features from multiple perspectives to predict overlapping points and mitigate the interference of non-overlapping points. Specifically, our approach utilizes pruning and weight-sharing quantization techniques to reduce model size and simplify the network structure. We evaluate the proposed model on noisy and partially overlapping point clouds from the ModelNet40 dataset, comparing its performance against other existing methods. Experimental results show that the proposed method significantly reduces the model's parameter size without compromising registration accuracy.

Human point cloud registration is a critical problem in the fields of computer vision and computer graphics applications. Currently, due to the presence of joint hinges and limb occlusions in human point clouds, point cloud alignment is challenging. To address these two limits, this paper proposes an unsupervised non-rigid human point cloud registration method based on deformation field fusion. The method mainly consists of the deep dynamic link deformation field estimation module and the probabilistic alignment deformation field estimation module. The deep dynamic link deformation field estimation module uses a time series network to convert non-rigid deformation into multiple rigid deformations. Then, feature extraction is performed to estimate the deformation field based on the rigid deformations. The probabilistic alignment deformation field estimation module builds on a Gaussian mixture model and adds local and global constraint conditions for deformation field estimation. Finally, the two deformation fields are fused into the total deformed field by aligning them, which enhances the sensitivity to both global and local feature information. The experimental results on public datasets and real private datasets demonstrate that the proposed method has higher accuracy and better robustness under joint hinges and limb adhesion conditions.

In recent years, CNN-Transformer hybrid networks, particularly those with dual-stream pyramid encoders, have gained significant attention in medical image analysis for their ability to combine the strengths of both CNNs and Transformers. However, a critical issue has been largely overlooked: the initial misalignment or inaccurate pre-alignment of image pairs, which can degrade the performance of these networks and limit their effectiveness on clinical datasets. To address this, we propose a novel multi-scale network framework for joint affine and deformable image registration tasks. Our approach introduces an additional branch to the dual-stream pyramid encoder, utilizing the Feature Selector to establish more precise spatial mappings for initially misaligned image pairs. This improves the network’s ability to correct large spatial misalignments during joint affine registration. Additionally, the Feature Translator in the decoder captures regions of interest based on image intensity, leveraging the inductive biases of convolutional layers to enhance pixel-level alignment. Our model demonstrates robust performance on registration tasks using two publicly available brain MRI datasets and is effective in addressing inter-frame mismatches in brain PET images. The code for our framework will be publicly available at https://github.com/ZZU-zy228/FSTNet.

Modern web applications use databases to store their data. When processing user requests, these applications retrieve and store data in the database server, which incurs network round trips. These round trips significantly increase the application's latency. Previous approaches have attempted to reduce these round trips by prefetching query results or batching database accesses. However, neither method can efficiently reduce the latency when some queries depend on previous queries' results. In real-world applications, nearly 50% of the queries depend on the result of other queries. This paper presents WeBridge, the first system capable of synthesizing stored procedures for large-scale real-world web applications. First, WeBridge employs concolic execution technique to analyze the applications and generate stored procedures for hot program paths. Then, it seamlessly integrates the stored procedures into the application by extending the database access library. Finally, it improves the efficiency of the stored procedures with speculative execution. Evaluation using real-world web applications and workloads show that WeBridge achieves up to 79.8% median latency reduction and up to 2× peak throughput.

Motivated by the fact that there exists the operation of conjugation in quantum systems, the concept of bicon-numbers is proposed in this article. The bicon-numbers are defined by introducing two symbolic parameters into the set of complex numbers. The basic functions of these two symbolic parameters are specified by an axiom which abstracts the operation of complex conjugation. Basic properties are developed for the operations of addition and multiplication in the bicon-number set. In addition, several different forms are given for bicon-numbers, and the corresponding operation rules are established. By exploring the relations of the vensors in the bicon-number set, the structure of the bicon-number set is depicted, and real matrix representations of bicon-numbers are also presented. Besides, bicomplex matrix representations for bicon-numbers are also investigated in view that the operation of multiplication for bicomplex numbers possesses commutativity property. In addition, the matrices with bicon-numbers as entries are investigated, and state responses of some quantum systems are given within the framework of bicon-numbers.

Today, most database-backed web applications depend on the database to handle deadlocks. At runtime, the database monitors the progress of transaction execution to detect deadlocks and abort affected transactions. However, this common detect-and-recover strategy is costly to performance as aborted transactions waste CPU resources.To avoid deadlock-induced performance degradation, developers aim to reorganize the application code to remove deadlocks. Unfortunately, doing so is difficult for web applications. Not only do their implementations include hundreds of thousands of LoCs, but they also use third-party object-relational mapping (ORM) frameworks which hide database access details. Consequently, it is hard for developers to accurately diagnose deadlocks.We propose WeSEER, a deadlock diagnosis tool for web applications. To overcome the opacity of ORMs, WeSEER performs concolic execution on unit tests to extract a web application’s transactions as a sequence of template statements with symbolic inputs as well as path conditions that enable the sequence. WeSEER then analyzes the extracted transactions based on fine-grained lock modeling to identify potential deadlocks and report the code locations that cause them. We implement WeSEER for Java-based (OpenJDK) web applications, and use it to analyze two popular open-source e-commerce applications, Broadleaf and Shopizer. WeSEER has successfully identified 18 potential deadlocks in Broadleaf and Shopizer. Eliminating these identified deadlocks can result in up to 39.5× and 4.5× throughput improvement for Broadleaf and Shopizer, respectively.

Distributed transaction systems incur extensive cross-node communication to execute and commit serializable OLTP transactions. As a result, their performance greatly suffers. Caching data at nodes that execute transactions can cut down remote reads. Batching transactions for validation and persistence can amortize the communication cost during committing. However, caching and batching can significantly increase the likelihood of conflicts, causing expensive aborts. In this paper, we develop Hackwrench to address the challenge of caching and batching. Instead of aborting conflicted transactions, Hackwrench tries to repair them using fine-grained re-execution by tracking the dependencies of operations among a batch of transactions. Tracked dependencies allow Hackwrench to selectively invalidate and re-execute only those operations necessary to "fix" the conflict, which is cheaper than aborting and executing an entire batch of transactions. Evaluations using TPC-C and other micro-benchmarks show that Hackwrench can outperform existing commercial and research systems including FoundationDB, Calvin, COCO, and Sundial under comparable settings.

We present XIndex, a concurrent ordered index designed for fast queries. Similar to a recent proposal of the learned index, XIndex uses learned models to optimize index efficiency. Comparing with the learned index, XIndex is able to effectively handle concurrent writes without affecting the query performance by leveraging fine-grained synchronization and a new compaction scheme, Two-Phase Compaction. Furthermore, XIndex adapts its structure according to run-time workload characteristics to support dynamic workload. We demonstrate the advantages of XIndex with both YCSB and TPC-C (KV), a TPC-C variant for key-value stores. XIndex achieves up to 3.2X and 4.4X performance improvement comparing with Masstree and Wormhole, respectively, on a 24-core machine, and it is open-sourced1.