Decentralized Federated Distillation (DFD) has emerged as a significant research direction since it not only supports heterogeneous model training but also naturally avoids privacy risks and communication bottlenecks stemming from the central server. DFD shows great potential in various training scenarios, especially in cross-silo federated environments. Existing DFD algorithms have limitations, including reliance on public datasets for distillation, insufficient privacy protection, and high communication overhead. This paper introduces Ring-Distill, a novel DFD algorithm designed specifically for cross-silo federated environments, effectively addressing the aforementioned limitations. Ring-Distill allows clients to complete federated training within a few global epochs (i.e., communication rounds) without sharing public datasets, e.g.,$N$global epochs for a system involving$N$clients. To protect privacy and further reduce communication overhead, Ring-Distill contains a privacy-oriented automatic model pruning mechanism (PAMP) which can automatically create the privacy-preserving compressed model called proxy model for each client. These proxy models are employed for client-to-client distillation during each global epoch. Furthermore, Ring-Distill contains a historical model-based distillation (HMD) mechanism that allows local models to transfer knowledge from multiple historical proxy model replicas stored locally. Due to historical proxy model replicas, the HMD mechanism not only improves the distillation performance but also effectively mitigates client dropout issues. Theoretical analysis and comprehensive experiments show that Ring-Distill has significant advantages in terms of accuracy, privacy, and communication cost.

GPU clusters have been widely used to co-locate various deep learning (DL) workloads in a multi-tenant way. Although such resource sharing can significantly reduce training cost, resource contention and interference among co-located workloads make task scheduling very complex and challenging. To simplify the scheduling problem, existing algorithms usually divide the procedure of scheduling into two sub-tasks, i.e., task placement and resource allocation, and allocate resources according to pre-defined and fixed resource demands. However, such a paradigm significantly constrains the selection of potential scheduling solutions. In this article, we present MAIFS, a novel multi-agent reinforcement learning based scheduling algorithm that handles task placement and resource allocation integratedly, and allows fungible resource allocation based on resource sensitivity of DL workloads. The core of MAIFS lies in two mechanisms. The multi-agent attention mechanism is designed to learn and share inter-related resource state features observed from different agents, which enables agents to explore fungible resource allocation solutions. The dynamic coordination graph mechanism is designed for coordinating interactive task placement decisions of agents during integrated scheduling, so as to mitigate potential task conflicts. Simulated experiments using two large scale production DL workload traces and physical deployment experiments based on a Kubernetes based GPU cluster show that MAIFS can outperform state-of-the-art scheduling algorithms by up to 44% in terms of makespan and 46% in terms of job completion time (JCT).

Federated Distillation (FD), a popular variant of Federated Learning (FL), has attracted researchers' attention due to its ability to support heterogeneous model training. Generally, FD allows clients to upload logits associated with public datasets for knowledge transfer, yet logits may pose privacy risks. In this study, we provide the first demonstration of the impact of privacy risks caused by logits. Specifically, we design a data reconstruction attack against logits named L-Attack which can reveal sensitive information about the target client without access to the target model. Via the zeroth-order optimization technique, L-Attack involves training a server-side generator that unveils certain features of private data owned by the target client. To defend against L-Attack, we propose a label aggregation-based FD algorithm called LabelAvg which allows clients to upload predicted hard labels for knowledge transfer instead of logits. Due to the insufficient information in labels for distillation, LabelAvg provides a voting-based label smoothing mechanism that enables the server to construct smooth labels from received labels. The generated smooth labels which stand for the consensus among all clients, indicate the approximate probability distribution. Thus, these smoothed labels bear a striking similarity to logits and can be used for distillation. Analysis and experimental results prove LabelAvg is superior to baselines in terms of accuracy, privacy, and communication data volume.

Workload prediction plays a crucial role in resource management of large scale cloud datacenters. Although quite a number of methods/algorithms have been proposed, long-term changes have not been explicitly identified and considered. Due to shifty user demands, workload re-locations, or other reasons, the “resource usage pattern” of a workload, which is usually quite stable in a short-term view, may change dynamically in a long-term range. Such long-term dynamic changes may cause significant accuracy degradation for prediction algorithms. How to handle such long-term dynamic changes is an open and challenging issue. In this article, we propose Evolution Graph for Workload Prediction (EvoGWP), a novel method that can predict long-term dynamic changes using a delicately designed graph-based evolution learning algorithm. EvoGWP automatically extracts shapelets to explicitly identify resource usage patterns of workloads in a fine-grained level, and predicts workload changes by considering factors in both temporal and spatial dimensions. We design a two-level importance based shapelet extraction mechanism to mine new usage pattern changes in temporal dimension, and design a novel evolution graph model to fuse the interference among resource usage patterns of different workloads in spatial dimension. By combining temporal extraction of shapelets from each single workload and spatial interference of shapelets among different workloads, we then design a spatio-temporal GNN-based encoder-decoder model to predict the long-term dynamic changes of workloads. Experiments using real trace data from Alibaba, Tencent and Google show that EvoGWP improves the prediction accuracy by up to 58.6% over the state-of-the-art prediction methods. Moreover, EvoGWP can outperform the state-of-the-art prediction methods in terms of model convergence. To the best of our knowledge, this is the first work that explicitly identifies fine-grained workload resource usage patterns to accurately predict long-term dynamic changes of workloads.

Function as a Service (FaaS) has become a popular computing paradigm. FaaS shifts the burden of resource management to cloud providers and enables users to focus only on the logic of the code. At the same time, FaaS has thrived due to its scalability and billing strategies. However, cold start has been always a major challenge for FaaS which causes additional delay. Existing cold start mitigation methods ignore dependencies among functions resulting in bloated function combinations and consequently unnecessary resource consumption. In this paper, we propose FaaSCom, a FaaS scheduler which mitigates cold start problem from the perspective of preallocating containers. We mine causality among functions and divide the function dependency graph into communities. By scheduling containers according to the function communities, we can reduce the number of containers provisioned for functions to decrease resource consumption while not increasing cold start occurrences. Besides, we design an extended histogram policy to mitigate cold starts for long-time interval functions. We evaluate our method on an industrial serverless dataset. Compared with baseline methods, FaaSCom reduces the cold start rate by 8.4% while saving 50% memory consumption.

Client selection (a.k.a., client sampling) is one of the hot topics in Federated Learning (FL). In each communication round, selecting some clients to participate in aggregation can effectively reduce the communication overhead caused by exchanging model parameters. However, due to statistical heterogeneity in FL, selecting clients randomly may affect the performance of aggregated global models. existing approaches regarding client selection firstly cluster clients and then sample(select) some representative clients from each cluster. However, these clustering-based approaches may be either time-intensive or high complexity. To address these issues, In this paper, we introduce Bi-level Sampling, a clustering-based approach for client selection. After multinomial distribution sampling, Bi-level Sampling clusters clients based on weighted per-label mean class scores and then selects participating clients for federated learning in each round. Bi-level Sampling can lead to better client representativity and the reduced variance of the client’s stochastic aggregation weights in FL. Our approach can be integrated into typical FL frameworks. Experimental results show that, compared with state-of-the-art approaches, our approach demonstrates significantly more stable and accurate convergence behavior-getting higher test accuracy and less training time, especially in highly Non-IID settings.

Co-location, which deploys long running applications and batch-processing applications in the same computing cluster, has become a promising way to improve resource utility for large cloud datacenters. However, co-location brings huge challenges to task scheduling because different types of workloads may affect each other. Existing works on task scheduling rarely focus on the scenario of co-location. This article presents Co-ScheRRL, a scheduling algorithm delicately designed for co-located workloads. Co-ScheRRL consists of two major mechanisms: i) a self-attention encoding mechanism which encodes and represents states of the computing cluster as a set of embedding feature vectors; ii) a deep reinforcement learning (DRL) relational reasoning mechanism which calculates and compares different scheduling actions under different co-located workloads pattern via DRL feedback reward signals based on these feature vectors. Our two mechanisms can tackle complicatedly and dynamically varying behaviors of co-located workloads. With the help of these two mechanisms, Co-ScheRRL is able to construct high-quality scheduling policies. Trace-driven simulation demonstrates that Co-ScheRRL outperforms existing scheduling algorithms in terms of makespan by more than 38.4% and throughput by more than 166.7%.

To efficiently generate safe trajectories for an autonomous vehicle in dynamic environments, a layered motion planning method with decoupled path and speed planning is widely used. This paper studies speed planning, which mainly deals with dynamic obstacle avoidance given a planned path. The main challenges lie in the optimization in a non-convex space and the trade-off between safety, comfort, and efficiency. First, this work proposes to conduct a search in second-order derivative space for generating a comfort-optimal reference trajectory. Second, by combining abstraction and refinement, an algorithm is proposed to construct a convex feasible space for optimization. Finally, a piecewise Bézier polynomial optimization approach with trapezoidal corridors is presented, which theoretically guarantees safety and significantly enlarges the solution space compared with the existing rectangular corridors-based approach. We validate the efficiency and effectiveness of the proposed approach in simulations.

The co-location of online services and offline tasks has become very popular in data centers, which can largely improve resource utilization. Scheduling co-located offline tasks is challenging due to the interference with online services. Existing co-location scheduling algorithms try to find the best combination of different workloads to avoid performance interference and maximize the utilization of data centers, but few of them take the time varying resource constraints into account. We propose a heuristic algorithm named interval matching scheduling algorithm based on the idea that the time series of available resources and task scheduling can be regarded as interval endpoints. The proposed scheduling algorithm makes decisions based on a scoring method that calculates the matching degrees of the tasks and the changing resource series. The experimental results show that the proposed algorithm has achieved better performance under different parameter settings comprehensively.

This paper investigates autonomous navigation for mobile robots with a low-cost monocular camera. The main challenge lies in: i) how to accurately detect prior unseen obstacles and acquire obstacles position from monocular images without depth information, and ii) how to get the constraints for path planning to generate safe and stable path for the robot. To accurately locate surrounding obstacles using only a monocular camera, we adopt a weakly-supervised semantic segmentation network trained from LIDAR data and perform inverse perspective transformation based on ground plane constraint. Meanwhile, to reduce segmentation noise, we establish a probability occupancy map based on the distance between robot and obstacles. For path generations, we present a novel search-and-optimization based planning approach to get boundary constraints in Frenet frame and generate stable local path with consecutive image inputs. In the simulation, segmentation Intersection over Union (IoU) of the drivable area achieves more than 99% and the average mapping accuracy is less than 10cm, showing feasibility and robustness of our navigation scheme.

The optimal control for mobile agents is an important and challenging issue. Recent work shows that using randomized mechanism in agents’ control can make the state unpredictable, and thus improve the security of agents. However, the unpredictable design is only considered in single period, which can lead to intolerable control performance in long time horizon. This paper aims at the trade-off between the control performance and state unpredictability of mobile agents in long time horizon. Utilizing random perturbations consistent with uniform distributions to maximize the attackers’ prediction errors of future states, we formulate the problem as a multi-period convex stochastic optimization problem and solve it through dynamic programming. Specifically, we design the optimal control strategy considering both unconstrained and input constrained systems. The analytical iterative expressions of the control are further provided. Simulation illustrates that the algorithm increases the prediction errors under Kalman filter while achieving the control performance requirements successfully.

This paper studies adaptive task allocation for multi-agent cooperation with unknown capabilities. The tasks considered here are single-type, large-scale and not prior known initially in every implementation. This scenario is quite common in numerous latest cooperative applications, like crowdsourcing. Since the amount of tasks are fixed, it is reasonable to assume the cost is constant, and the consumed time becomes an important index. To minimize the cost time, the main challenges lie in how to allocate tasks to complete these tasks in a decentralized way and avoid solving a new optimization problem at each step. The advantages and novelty of our work are threefold: i) Leveraging the consensus and distributed method, we transform the minimum-time problem into a solvable distributed optimization problem. By consensus algorithm, the allocation method is expressed explicitly and easily. ii) We prove that the convergence of assignment process and optimal allocation is achieved geometrically. iii) The proposed algorithm is extended to the case where the efficiencies of agents are stochastic, and simulations demonstrate the effectiveness of our approach.