All-flash servers, while being widely popular for their high performance and large capacity, can incur significant energy consumption in modern storage systems. Through a motivational study, we discover that the culprit is the inefficiency in the software stack, and existing power-saving methods fail to deliver comparable performance, especially under workload bursts. Guided by the lessons learned, we propose Sandman, a scheduling framework that combines the fast resource scaling mechanism, resource monitoring, and I/O burst detection policies. Experiments show that Sandman reduces average power consumption by up to 39.38% and energy consumption by up to 33.36% while delivering performance comparable (within 5% in corner cases) to the best performance case (the busy-polling stack) in both benchmarks and field workloads.

Cloud local disks are attractive for their affordable price and high performance. The recent advancement in CPUs motivates cloud vendors to further multiplex the computing resources to serve more users. Unfortunately, such proposals are constrained by the limited offerings of cloud local disks per server as the underlying storage devices are either large but slow (e.g., HDDs) or fast yet small (e.g., NVMe SSDs).

GNN-based social recommendation aims to use social network information to improve recommendation performance of traditional user–item interaction network (U–I network). However, in graph neural network information aggregation, both social networks and U–I networks inevitably have noise, which affects accuracy of recommendation results. To reduce the noise impact of network data, we propose Robust Social Recommendation based on Contrastive Learning and Dual-Stage Graph Neural Network (CLDS). First, considering instability of social networks, we propose the social preference network. It is robust and retains only social friend relationships with common preferences. Based on it and U–I network, we construct a social recommendation pre-training model. Next, we propose self-contrastive learning method. The method initializes multiple social network node representations through Gaussian distribution , pre-training and random disturbance, respectively. Then, it uses contrastive learning on the generated multiple node representations to enhance the robustness of node representation. Finally, CLDS avoids directly capturing potentially user–user and item–item information in U–I networks which is incomplete and untrusted. And instead, it only extracts user–item information to reduce the noise generated by GNN-based U–I network information aggregation. We conduct experiments under the open-source real network dataset. The experimental results show that CLDS outperforms state-of-art methods in social recommendation. The code is available at: https://github.com/Andrewsama/CLDS-master .

The high-precision clock synchronization is pursued with the consideration of the balance for time-domain resource utilization, when 5G technologies are expected to carry the timing services. Firstly, a clock synchronization model is established in the case that the system exists the observed value loss. The probability distribution of the loss of the observed value is evaluated by a newly proposed delay deterministic confidence method. Then, the error boundness of the clock synchronization is investigated by the Kalman filter algorithm, and the optimization problem for 5G time-domain resource configuration is formulated with guaranteeing the precision of clock synchronization; Finally, the proposed optimization problem is solved by dueling double deep Q-learning-based Lyapunov optimization. The experimental results verify the effectiveness and superiority of the proposed method in terms of the joint optimization of clock synchronization error covariance and throughput.

More and more Non-Volatile Memories (NVM) have been widely applied to various embedded systems to build the heterogeneous memory architecture. However, the write-endurance of NVM remains a great challenge. Hence, we should take full consideration of the write-endurance of NVM when allocating data on heterogeneous memory architecture. There is an observation that, for most real workloads, about 10% of data account for 90% write operations. This brings us an opportunity to reduce the write wear of NVM through carefully allocating write-intensive data. In this paper, we explore the problem that how to find a balance between the system cost and write-endurance of NVM for data allocation on heterogeneous memory architecture. We propose a write-aware data allocation algorithm, WADA. WADA can not only greatly reduce the write wear of NVM, but also guarantee the near-optimal system cost. We also propose an integer linear programming (ILP) model to generate an optimal data allocation, which can obtain the minimum cost. The result of ILP can be used as a standard to evaluate the efficiency of other algorithms. Experiments show that WADA outperforms all the other algorithms on both system cost and write wear of NVM. Compared to previous algorithms, WADA can reduce up to 47.77% system cost and 60.89% write wear of NVM. Compared to ILP, WADA can achieve the near-optimal system cost within just 2% difference.

Mutual information (MI), as a similarity measure has been very successful in multi-modal image registration. However, the iterative nature of the method for finding optimal transformation makes it time consuming and mis-registration can happen due to local extrema. In this paper, a heuristic approach based on straight line detection and matching is proposed for rigid registration to speed up mutual information-based method and avoid local extrema by converting the 3-D optimization problem into a 1-D search along the corresponding lines. The proposed method is suitable for images rich with line segments, such as those used in remote sensing applications.