Nowadays, more and sensors, devices applications are connected in Industrial Internet of Things (IIoT), producing massive real-time flows which need to be scheduled for Quality-of-Service provision. To realize application-aware adaptive flow scheduling, the problem traffic classification must addressed at first. When edge computing paradigm is introduced into IIoT, service can deployed on node near-end. Recently, deep-learning-based IIoT methods show better performance, but computational...

The rise of Machine Learning as a Service (MLaaS) has driven the demand for complex and customized real-time inference tasks, often requiring cascading multiple deep neural network (DNN) models into pipelines. However, these pipelines pose significant challenges due to scheduling complexity, particularly in maintaining strict latency service level objectives (SLOs). Existing systems serve with model-independent policies, which ignore unique workload characteristics introduced by model...

Load balancing is essential for datacenter networks. However, prior solutions have significant limitations: they either are oblivious to congestion or involve a daunting and time-consuming parameter-tunning task over their heuristics achieving good performance. Thus, we ask: it possible learn balance traffic? While deep reinforcement learning (DRL) sounds like answer, observe that too heavyweight due the long decision-making latency. Therefore, introduce BULB, lightweight automated load...

Abstract A series of perfluorocyclobutyl (PFCB) aryl ether‐based amphiphilic diblock copolymers containing hydrophilic poly(acrylic acid) (PAA) and fluorophilic poly( p ‐(2‐( ‐tolyloxy)perfluorocyclobutoxy)phenyl methacrylate) segments were synthesized via successive atom transfer radical polymerization (ATRP). 2‐MBP‐initiated CuBr/ N,N,N ′ ,N ″‐pentamethyldiethylenetriamine‐catalyzed ATRP homopolymerization the PFCB‐containing methacrylate monomer, methacrylate, can be performed in a...

Over the last decade, we have witnessed growing data volumes generated and stored across geographically distributed datacenters. Processing such geo-distributed datasets may suffer from significant slowdown as underlying network flows to go through inter-datacenter networks with relatively low highly heterogeneous available link bandwidth. Thus, optimizing transmissions of flows, especially coflows that capture application-level semantics, is important for improving communication performance...

Low latency stream processing on large clusters consisting of hundreds to thousands servers is an increasingly important challenge. A crucial barrier tackling this challenge stragglers, i.e., tasks that are significantly straggling behind others in the data. However, prior straggler mitigation solutions have significant limitations. They balance streaming workloads among but may incur imbalanced backlogs when exhibit variance, causing stragglers as well. Fortunately, we observe carefully...

The coexistence of RDMA and TCP is prevalent in the datacenter. Despite sound isolation at end hosts, they share same switches network. Their different networking behaviors (E.g., hardware demand transport protocols) lead to huge differentiated buffer for switches. However, existing management schemes ignore these dissimilarities simply treat such RDMA/TCP mix-flows as typical multi-class traffic, resulting inferior degrading performances. This paper presents BRT, a first systematic solution...

Distributed machine learning (DML) is an increasingly important workload. In a DML job, each communication phase can comprise <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">coflow</i> , and there are dependencies among its coflows. Thus, efficient coflow scheduling becomes critical for jobs. However, the majority of existing solutions focus on single-stage coflows with no dependencies. While few studies schedule dependent multi-stage jobs,...

Container overlay network, though being widely adopted to enable communication between containers on different hosts, is a key downside for latency-sensitive applications. The state-of-the-art solution seeks shorten the data path in packet processing by replacing connection file descriptors with host namespace ones. While promising, it must block each until relevant set up, thus heavily influencing request latency. In this paper, we present ShuntFlow, systematic delivery framework that...

Keyword extraction has been a very traditional topic in Natural Language Processing. However, most methods have too complicated and slow to be applied real applications, for example web-based system. This paper proposes an approach which will complete some preparing works focusing on exploring the linguistic characteristics of specific domain. part can completed once all thus reduce burden process. It is weighted sum method work focus finding out weight. Once we weight, by addition,...

Distributed streaming applications require the underlying network flows to transmit packets continuously keep their output results fresh. These will become stale if no updates come, and staleness is determined by slowest flow. At this point, coflows can be semantically comprised. Hence, efficient coflow transmission critical for applications. However, prior coflow-based solutions have significant limitations. They use a one-shot performance metric-CCT (coflow completion time), which cannot...

Data-parallel applications generate a mix of coflows with and without deadlines. Deadline are mission-critical must be completed within deadlines, while the non-deadline desire to as soon possible. Scheduling such mix-coflows is an important problem in modern datacenters. However, existing solutions only focus on one two types coflows: they either solely concentrate meeting deadlines deadline-aware or reducing coflow completion times (CCTs) coflows. In this article, we study optimizing...

Unified modeling language (UML) activity diagrams depict the internal behavior of different program operations with help nodes and edges, describing business logic in user requirements. Traditionally, requirements engineers practitioners refer to process documents analyze them build UML manually, which makes labor time consuming. Recently, deep learning technology has been utilized various fields achieved excellent results. We propose a novel pipeline, named TAG, for automatically generating...

Data-parallel applications generate a mix of coflows with and without deadlines. Deadline are mission-critical must be completed within deadlines, while non-deadline desire to as soon possible. Scheduling such mix-coflows is an important problem in modern datacenters. However, existing solutions only focus on one the two types coflows: they either solely meeting deadlines deadline-aware or reducing coflow completion times (CCTs) coflows. In this paper, we study optimizing deadline...

Solar simulator is a key instrument for photovoltaic field, which aims to act the role of natural sunlight irradiance indoor, and we should identify how similar they are in quantity. The critical factor similarity lies its spectral irradiance, because solar cells' wavelength-dependent responsivity, mismatch can induce large errors during characteristic parameters measurement. In this article, method measuring simulator's was proposed along with uncertainty analysis. A calibrated fiber optic...