Function-as-a-Service (FaaS) is a promising cloud computing model known for its scalability and elasticity. In various application domains, FaaS workflows have been widely adopted to manage user requests complete computational tasks efficiently. Motivated by the fact that function containers collaboratively use image layer's memory, co-placing functions would leverage memory sharing reduce cluster footprint, this paper studies layer- wise serverless functions. We find overwhelming placing in...

With the rapid evolution of Internet, Internet Things (IoT), and geographic information systems (GIS), spatio-temporal Big Data (STBD) is experiencing exponential growth, marking onset STBD era. Recent studies have concentrated on developing algorithms techniques for collection, management, storage, processing, analysis, visualization STBD. Researchers made significant advancements by enhancing handling techniques, creating novel systems, integrating support into existing systems. However,...

In serverless computing, cold start results in long response latency. Existing approaches strive to alleviate the issue by reducing number of starts. However, our measurement based on real-world production traces shows that minimum starts does not equate latency, and solely focusing optimizing will lead sub-optimal performance. The root cause is functions have different priorities terms latency benefits transferring a warm start. this paper, we propose <italic...

With the rapid development of Internet Things (IoT), IoT devices find applications in various domains. The data generated by these is utilized for analysis and services, especially field Artificial Intelligence (AI) applied to IoT, known as (AIoT). enhancement edge device computing power has led emergence research areas like edge-cloud synergy AI theories application services. In context lifelong learning real-time processes AIoT addressing unseen tasks becomes crucial. Unseen arise when...

Sharding, breaking nodes into smaller groups, aims to enhance the scalability of traditional blockchain systems by allowing parallel transaction processing. However, existing sharding methods face challenges, including heavy inter-shard communication, re-sharding overhead, and low consensus concurrency. These limitations ultimately result in less desired system performance. To address these we propose Frustum, a novel hierarchical pipelined system. It separates shards two layers: top L-Shard...

Anycast routing is very useful for many applications such as resource discovery in Delay Tolerant Networks (DTNs).In this paper, based on a new DTN model, we first analyze the anycast semantics DTNs.Then present novel metric named EMDDA (Expected Multi-Destination Anycast) and corresponding algorithm DTNs.Extensive simulation results show that proposed scheme can effectively improve efficiency of DTNs.It outperforms another algorithm, Minimum Expected (MED) by 11.3% average term delays 19.2%...

Cloud-based file systems are widely accepted and adopted for personal business purposes in recent years. Statistics shows that ∼25% of operations from a typical user random writes. Inherited traditional disk-based systems, most distributed also based on objects or chunks fixed sizes, which work well sequential writes but poorly This paper investigates the design paradigm variable-sized system, where new write interface is proposed to provide rich semantics. A novel system named VarFS,...

As widely used indices in key-value stores, the Log-Structured Merge-tree (LSM-tree) and its variants suffer from severe write amplification due to frequent merges compactions for write-intensive applications. To address problem, we first propose Append-tree (LSA-tree), which tries compact data with appends instead of merges, significantly reduces solves issues existed current append trees. However LSA increases read space amplifications. Furthermore based on LSA, design Integrated...

Generative Adversarial Networks (GANs) have become predominant in mobile computing for their ability to generate data. The concern data privacy has made it arduous collect large-scale datasets GAN training on centralized servers. Federated Learning (FL) emerged as a promising solution address concerns. In this paper, we propose Oasis, multiplayer-oriented federated system. We present motivation, highlighting the Nash Equilibrium (NE) shift vanilla GANs, exacerbated by heterogeneity, leading...

In this paper, we explicitly consider the scenario of supporting applications with high-bandwidth and low-latency requirements in mobile social networks (MSNs), by leveraging existing network sites. Our solution does not assume any centralized server who coordinates data storage, access, group management. We address fundamental challenge such applications, namely: limited bandwidth availability intense requirement that may hinder deployment these applications. propose a link-based congestion...

Internet of Things (IoT) gateways integrate various sensors and compute initial decisions before transmitting data to the cloud for further processing. As functions they need support become increasingly complex, must upgrade their hardware. Network (NF) video analytics (VAs) are two typical examples hardware requirements: NFs specialized accelerators, while VAs parallel processing power. However, typically constrained by factors, such as power, size, cost, leading a multiplex minimize...

Existing decentralized learning of Generative Adversarial Network (GAN) suffers from a slower convergence rate and training instability due to the changes in gradient-sharing approach among workers. It requires more iterations achieve deteriorates GAN training, consequently leading slow down speed accuracy degradation. We propose JediGAN, novel distributed system that achieves optimal benefits by balancing communication overheads through adaptive scheduling strategies. JediGAN reduces...