The availability of high-speed solid-state storage has introduced a new tier into the hierarchy. Low-latency and high-IOPS drives (SSDs) cache data in front high-capacity disks. However, most existing SSDs are designed to be drop-in disk replacement, hence mismatched for use as cache.

Next-generation non-volatile memories (NVMs) promise DRAM-like performance, persistence, and high density. They can attach directly to processors form main memory (NVMM) offer the opportunity build very low-latency storage systems. These high-performance systems would be especially useful in large-scale data center environments where reliability availability are critical. However, providing NVMM is challenging, since latency of replication overwhelm low that should provide. We propose Mojim,...

The monolithic server model where a is the unit of deployment, operation, and failure meeting its limits in face several recent hardware application trends. To improve resource utilization, elasticity, heterogeneity, handling datacenters, we believe that datacenters should break servers into disaggregated, network-attached components. Despite promising benefits disaggregation, no existing OSes or software systems can properly manage it.We propose new OS called splitkernel to disaggregated...

Recently, there is an increasing interest in building data-center applications with RDMA because of its low-latency, high-throughput, and low-CPU-utilization benefits. However, not readily suitable for datacenter applications. It lacks a flexible, high-level abstraction; performance does scale; it provide resource sharing or flexible protection. Because these issues, difficult to build RDMA-based exploit RDMA's

Attaching next-generation non-volatile memories (NVMs) to the main memory bus provides low-latency, byte-addressable access persistent data that should significantly improve performance for a wide range of storage-intensive workloads. We present an analysis storage application with (NVMM) using hardware NVMM emulator allows fine-grain tuning parameters. Our evaluation results show improves over flash-based SSDs and HDDs. also compare applications running on realistic same idealized as DRAM....

Utilizing Internet of Things (IoT) technology in smart grid is an important approach to speed up the informatization power system, and it beneficial for effective management infrastructure. Disaster prevention reduction transmission line one most application fields IoT. Advanced sensing communication technologies IoT can effectively avoid or reduce damage natural disasters lines, improve reliability economic loss. Focused on characteristic construction development grid, this paper introduced...

Memory disaggregation has attracted great attention recently because of its benefits in efficient memory utilization and ease management. So far, research all taken one two approaches: building/emulating nodes using regular servers or building them raw devices with no processing power. The former incurs higher monetary cost faces tail latency scalability limitations, while the latter introduces performance, security, management problems.

We present Nameless Writes, a new device interface that removes the need for indirection in modern solid-state storage devices (SSDs). writes allow to choose location of write; only then is client informed name (i.e., address) where block now resides. Doing so allows control block-allocation decisions, thus enabling it execute critical tasks such as garbage collection and wear leveling, while removing large costly tables. demonstrate effectiveness nameless by porting Linux ext3 file system...

Rust is a young programming language designed for systems software development. It aims to provide safety guarantees like high-level languages and performance efficiency low-level languages. The core design of set strict rules enforced by compile-time checking. To support more controls, allows programmers bypass these compiler checks write unsafe code.

Go is a statically-typed programming language that aims to provide simple, efficient, and safe way build multi-threaded software. Since its creation in 2009, has matured gained significant adoption production open-source advocates for the usage of message passing as means inter-thread communication provides several new concurrency mechanisms libraries ease multi-threading programming. It important understand implication these proposals comparison shared memory synchronization terms program...

To achieve automatic, fast, efficient and high-precision pavement distress classification detection, road surface image detection models based on deep learning are trained. First, a dataset is built, including 9017 pictures with distress, 9620 without distress. These were captured from 4 asphalt highways of 3 provinces in China. In each image, there exists one or more types alligator crack, longitudinal block transverse pothole patch. The distresses labeled by rectangle bounding box the...

With the development of mobile platform, such as smart cellphone and pad, E-Learning model has been rapidly developed. However, due to low completion rate for it is very necessary analyze behavior characteristics online learners intelligently adjust education strategy enhance quality learning. In this paper, we analyzed relation indicators build student profile gave countermeasures. Adopting similarity computation Jaccard coefficient algorithm, designed a system clean dig into educational...

Mobile edge computing (MEC) in cyber-physical systems (CPSs) with massive resource-constrained node (ECN) faces new challenges security provisioning. The traditional centralized authentication schemes low performance are no longer applied for MEC CPS. Due to the mobility of ECN, it is extraordinarily practical ECN establish a association another AP once leaving service area its current AP. In this paper, we represent related research and propose novel efficient software-defined networking...

RDMA technology enables a host to access the memory of remote without involving CPU, improving performance distributed in-memory storage systems. Previous studies argued that suffers from scalability issues, because NIC's limited resources are unable simultaneously cache state all concurrent network streams. These concerns led various software-based proposals reduce size this by trading off performance.

Today's gen-AI workflows that involve multiple ML model calls, tool/API data retrieval, or generic code execution are often tuned manually in an ad-hoc way is both time-consuming and error-prone. In this paper, we propose a systematic approach for automatically tuning workflows. Our key insight can benefit from structure, operator, prompt changes, but unique properties of require new optimization techniques. We AdaSeek, adaptive hierarchical search algorithm autotuning AdaSeek organizes...

Large caches in storage servers have become essential for meeting service levels required by applications. These need to be warmed with data often today due various scenarios including dynamic creation of cache space and server restarts that clear contents. When large are at the rate application I/O, warmup can take hours or even days, thus affecting both performance load over a long period time.We created Bonfire, mechanism accelerating warmup. Bonfire monitors workloads, logs important...

Anesthesia and/or surgery may promote Alzheimer's disease (AD) by accelerating its neuropathogenesis. Other studies showed different findings. However, the potential sex difference among these has not been well considered, and it is unknown whether male or female AD patients are more vulnerable to develop postoperative cognitive dysfunction. We therefore set out perform a proof of concept study determine anesthesia can have effects in transgenic (Tg) mice, Tg plus Cyclophilin D knockout...

Energy saving and emission reduction have become common concerns in countries around the world. In China, with implementation of new strategy “carbon peak neutrality” rapid development smart grid infrastructure, amount data actual power dispatching fault analysis show exponential growth, which has led to phenomena such as poor supervision effectiveness difficulty handling faults process operation maintenance. Existing research on retrieval recommendation methods had a lower accuracy rate at...

Network intrusion data are characterized by high feature dimensionality, extreme category imbalance, and complex nonlinear relationships between features categories. The actual detection accuracy of existing supervised intrusion-detection models performs poorly. To address this problem, paper proposes a multi-channel contrastive learning network-based method (MCLDM), which combines in the stage extraction unsupervised to train an effective model. objective is research whether enrichment use...

Due to the vulnerable environment, limited recourse and open communication channel, wireless sensor networks (WSNs) are necessary be protected from various attacks. The key management is an important way protect in WSNs. In this paper, we present a hierarchical scheme (HKMS) which can efficiently enhance security survivability for clustered Different previous works, HKMS distributes keys based on hop counts one-way function by architecture, not only localizes things but also has no overhead....