Cloudless-Training:基于serverless的高效跨地域分布式ML训练框架

跨地域分布式机器学习(ML)训练能够联合多区域的云资源协作训练,可满足许多新兴ML场景(比如大型模型训练、联邦学习)的训练需求。但其训练效率仍受2方面挑战的制约。首先,多区域云资源缺乏有效的弹性调度,这会影响训练的资源利用率和性能;其次,模型跨地域同步需要在广域网(WAN)上高频通信,受WAN的低带宽和高波动的影响,会产生巨大通信开销。本文提出Cloudless-Training,从3个方面实现高效的跨地域分布式ML训练。首先,它基于serverless计算模式实现,使用控制层和训练执行层的2层架构,支持多云区域的弹性调度和通信。其次,它提供一种弹性调度策略,根据可用云资源的异构性和训练数据集的分布自适应地部署训练工作流。最后,它提供了2种高效的跨云同步策略,包括基于梯度累积的异步随机梯度下降(ASGD-GA)和跨云参数服务器(PS)间的模型平均(MA)。Cloudless-Training是基于OpenFaaS实现的,并被部署在腾讯云上评估,实验结果表明Cloudless-Training可显著地提高跨地域分布式ML训练的资源利用率(训练成本降低了9.2%~24.0%)和同步效率(训练速度最多比基线快1.7倍),并能保证模型的收敛精度。

基于Serverless架构的人工智能实验平台的设计与实现

针对实验环境搭建复杂、实验数据量大以及算法实现难度高等人工智能实验教学中的现实问题,设计并实现了基于Serverless架构的人工智能实验平台。通过整合容器构建、工作负载管理以及事件触发这三者来完成Serverless架构的设计与搭建;平台使用Docker image作为实验容器镜像,镜像中集成scikit-learn算法库与大量公用数据集完成海量人工智能基础算法的迁移;实验教学管理员以提交实验代码包的形式新增实验,学生可以使用实验容器中大量公用算法完成实验,提高了平台的可扩展性和易用性;同时,平台基于Serverless架构实现实验容器的动态扩缩容,增加了平台的多用户并发能力。

ATSSC:An Attack Tolerant System in Serverless Computing

Serverless computing is a promising paradigm in cloud computing that greatly simplifies cloud programming.With serverless computing,developers only provide function code to serverless platform,and these functions are invoked by its driven events.Nonetheless,security threats in serverless computing such as vulnerability-based security threats have become the pain point hindering its wide adoption.The ideas in proactive defense such as redundancy,diversity and dynamic provide promising approaches to protect against cyberattacks.However,these security technologies are mostly applied to serverless platform based on“stacked”mode,as they are designed independent with serverless computing.The lack of security consideration in the initial design makes it especially challenging to achieve the all life cycle protection for serverless application with limited cost.In this paper,we present ATSSC,a proactive defense enabled attack tolerant serverless platform.ATSSC integrates the characteristic of redundancy,diversity and dynamic into serverless seamless to achieve high-level security and efficiency.Specifically,ATSSC constructs multiple diverse function replicas to process the driven events and performs cross-validation to verify the results.In order to create diverse function replicas,both software diversity and environment diversity are adopted.Furthermore,a dynamic function refresh strategy is proposed to keep the clean state of serverless functions.We implement ATSSC based on Kubernetes and Knative.Analysis and experimental results demonstrate that ATSSC can effectively protect serverless computing against cyberattacks with acceptable costs.

A Secure and Cost-Effective Training Framework Atop Serverless Computing for Object Detection in Blasting

The data analysis of blasting sites has always been the research goal of relevant researchers.The rise of mobile blasting robots has aroused many researchers’interest in machine learning methods for target detection in the field of blasting.Serverless Computing can provide a variety of computing services for people without hardware foundations and rich software development experience,which has aroused people’s interest in how to use it in the field ofmachine learning.In this paper,we design a distributedmachine learning training application based on the AWS Lambda platform.Based on data parallelism,the data aggregation and training synchronization in Function as a Service(FaaS)are effectively realized.It also encrypts the data set,effectively reducing the risk of data leakage.We rent a cloud server and a Lambda,and then we conduct experiments to evaluate our applications.Our results indicate the effectiveness,rapidity,and economy of distributed training on FaaS.

基于Serverless架构的知识管理平台架构设计

针对传统知识管理平台架构日益复杂、难以跟随企业需求变化而快速演进的问题,提出了一种基于无服务器(Serverless)架构的知识管理平台的架构设计,充分利用云服务厂商提供的函数计算和事件总线等云服务,极大地降低了平台建设和维护的工作量,同时具有松耦合、高弹性、高可用性和高数据持久性等特点,可以满足不同规模企业对知识管理平台的需求。

基于演化博弈的Serverless移动目标防御决策方法

“无服务器”(Serverless)是一种新型计算范式,具有轻量、敏捷的特点,由于其架构的特殊性,导致在原有云安全问题的基础上又引入一些新的安全威胁。针对在未知攻防场景下难以准确选择最优化防御策略的问题,从虚拟化层和应用层两个方面根据移动目标防御(Moving Target Defense,MTD)多样化的思想提出防御策略。将具有玻尔兹曼探索的Q-Learning算法与复制动态方程结合,从有限理性的角度出发构建具有探索机制的演化博弈模型,使防御者在反复的攻防对抗中不断进行试错—探索—实施,最终获得最优化防御策略,获取最大收益。实验表明,引入探索机制的演化博弈模型具有可预测性,且演化博弈均衡点具有较强的稳定性。

HCloud:A Trusted JointCloud Serverless Platform for IoT Systems with Blockchain

Cloud computing has been exploited in managing large-scale IoT systems.IoT cloud servers usually handle a large number of requests from various IoT devices.Due to the fluctuant and heavy workload,the servers require the cloud to provide high scalability,stable performance,low price and necessary functionalities.However,traditional clouds usually offer computing service with the abstraction of virtual machine(VM),which can hardly meet these requirements.Meanwhile,different cloud vendors provide different performance stabilities and price models,which fluctuate according to the dynamic workload.A single cloud cannot satisfy all the requirements of the IoT scenario well.The JointCloud computing model empowers the cooperation among multiple public clouds.However,it is still difficult to dynamically schedule the workload on different clouds based on the VM abstraction.This paper introduces HCloud,a trusted JointCloud platform for IoT systems using serverless computing model.HCloud allows an IoT server to be implemented with multiple serverless functions and schedules these functions on different clouds based on a schedule policy.The policy is specified by the client and includes the required functionalities,execution resources,latency,price and so on.HCloud collects the status of each cloud and dispatches serverless functions to the most suitable cloud based on the schedule policy.By leveraging the blockchain technology,we further enforce that our system can neither fake the cloud status nor wrongly dispatch the target functions.We have implemented a prototype of HCloud and evaluated it by simulating multiple cloud providers.The evaluation results show that HCloud can greatly improve the performance of serverless workloads with negligible costs.

基于Serverless的反溯源技术应用研究

随着网络逐渐成为意识形态较量的主战场,攻防双方的技术手段在不断博弈中日渐精进,现有的反溯源手段无法避免防守方多维多技术的溯源手段,更易被防守方溯源反制。该文提出了一种基于Serverless的反溯源技术应用思路,利用Serverless的事件驱动和自动伸缩特性,使得用户在请求目标时,自动调用不同可用区域的IP地址,以此达到隐藏自身真实IP的目的。同时,由于Serverless实现应用开发与服务器分离,攻击者可直接进行攻击代码编写,也更加利于隐藏身份。通过利用Serverless中的云函数和CobaltStrike软件进行试验验证其可行性,发现其能很好地隐藏攻击源,防守方无法溯源到真实的攻击源。同时从防守方角度,详细分析流量特征,基于特征值和访问统计特征两个维度,构建攻击检测模型。通过模拟实际攻击行为和正常业务行为,验证了检测模型能够很好地发现攻击行为,并能区分攻击行为和正常业务行为,在一定程度上可以减少误报,降低对正常业务的影响,提高安全事件的处置效率,为防守方的入侵检测提供了检测思路。

一种高性能Serverless计算设计

FaaS(Function as a Service)允许开发者在不管理服务器的情况下在云端上传和执行代码。为解决使用FaaS过程中出现的性能和扩展问题,提出并证明了一种轻量级高性能Serverless设计,该设计提供优于现有解决方案的性能和扩展优势,并且可以实现细粒度的即用即付模型。比较了相同算法下的Docker和WebAssembly实现之间使用的性能基准。还讨论了在Serverless计算中采用WebAssembly的存在问题。

基于Serverless的二维码销售应用的设计与实现

我国在二维码应用技术领域处于领先地位,但目前二维码行业缺乏统一行业标准,服务过程没有专业的监督,尤其是一物一码的二维码落地应用尚不多见。本文提出的基于Serverless的二维码销售应用,能够实现一物一码的功能,用户扫描商品的二维码即可购买,增加复购,经销渠道独立拥有。鉴于微信小程序具有速度快、可分享、低门槛等优点,本文用户端应用微信小程序。小程序的开发技术采用基于Serverless云开发技术,无需搭建服务器,能够节省大量的时间和开发成本,从而提高利润。

Serverless上云 腾讯云的“三板斧”

企业的应用开发也能三步上云,你相信吗?今天,在Serverless技术的支持下,腾讯云确实可以做到协助企业三步上云。下面,就让我们来一探究竟。披萨饼店与Serverless的故事Serverless为什么会一夜爆红?先来看个披萨饼店的传奇。成立于2009年的深圳市乐凯撒比萨餐饮管理有限公司(以下简称乐凯撒),以其首创的榴莲披萨在短短几年时间内风靡全国,在国内多个城市开设了连锁餐厅。

Exploring serverless computing for stream analytic

This work proposes ARS(FaaS) serverless framework scheduling and provisioning resources for streaming applications autonomously, which ensures real-time response on unpredictable and fluctuating streaming data. A HPC cloud platform is used as a de facto platform, on which serverless computing for stream analytic is explored. This work enables application developers to build and run steaming applications without worrying about servers, which means that the developers are able to focus on application features instead of scheduling and provisioning resources of the infrastructure. The serverless computing framework, ARS(FaaS), provides function-as-a-service to make the developers write code in discrete event-driven functions. ARS(FaaS) is capable of running and scaling the developer's code automatically, according to the throughput of streaming events. The major contribution of this serverless framework is effective and efficient autonomous resource scheduling for real-time streaming analytic, which enables the developers to build applications faster with autonomous resource scheduling. ARS(FaaS) framework is appropriate for real-time and stream analytic on event-driven data with spiky and variable compute requirements.

Result-as-a-Service (RaaS): Persistent Helper Functions in a Serverless Offering

Serverless Computing or Functions-as-a-Service (FaaS) is an execution model for cloud computing environments where the cloud provider executes a piece of code (a function) by dynamically allocating resources. When a function has not been executed for a long time or is being executed for the first time, a new container has to be created, and the execution environment has to be initialized resulting in a cold start. Cold start can result in a higher latency. We propose a new computing and execution model for cloud environments called Result-as-a-Service (RaaS), which aims to reduce the computational cost and overhead while achieving high availability. In between successive calls to a function, a persistent function can help in successive calls by precomputing the functions for different possible arguments and then distributing the results when a matching function call is found.

浅析基于Serverless服务架构的应用开发

Serverless架构近几年频繁出现在一些技术架构大会的演讲标题中。很多人对于Serverless架构,仅从字面意义上理解,认为是一种无服务器架构,实际上它真正的含义是开发者再也不用过多考虑服务器的问题,它不是完全去除服务器,而是依靠第三方资源服务器后端。Serverless架构依托容器技术,提供了高并发、高兼容的特性,为开发者隐藏了底层服务器的细节,同时采取了更经济的按调用次数或时间计费的服务运营模式,因而备受关注。文章首先介绍Serverless架构的概念、组成以及特点,然后分析了Severless架构前后端一体化研发模式,随后列举了Severless架构适用场景的实践,最后阐述了Severless架构的技术缺点、当前面临的挑战以及未来的展望。

基于Serverless的区块链系统设计

由于区块链应用的发展常常受到性能、存储瓶颈和技术门槛的制约,本文提出一种基于Serverless的区块链系统:通过在区块链中引入分布式哈希表和Serverless微服务云存储,保障区块链存储的性能和安全性;通过函数计算服务进行链下协同计算,实现区块链事务分类处理,大幅降低用户成本的同时提高了系统的效率;提出Serverless框架下的智能合约开发和部署模式,相比传统DAPP开发有效降低技术门槛和维护成本。为利用Serverless技术解决区块链安全、性能、存储和合约开发问题提供方案。

A Designated Query Protocol for Serverless Mobile RFID Systems with Reader and Tag Privacy

Recently, a new type of Radio Frequency IDentification （RFID） system with mobile readers is introduced. In such a system, it is more desirable for mobile readers to identify tags without a back-end server, and thus it is frequently referred as a serverless mobile RFID system. In this paper, we formalize a serverless mobile RFID system model and propose a new encryption-based system that preserves the privacy of both tags and readers in the model. In addition, we define a new adversary model for the system model and show the security of the proposed system. Throughout comparisons between ours and the other alternatives, we show that our proposed system provides a stronger reader privacy and robustness against a reader forgery attack than the competitors.

Implementation of serverless cloud GIS platform for land valuation

Cloud computing enables performing computations and analysis tasks and sharing services in web-based computer centres instead of local desktop systems.One of the most used areas of cloud computing is geographic information systems(GIS)applications.Although Desktop GIS products are still used in the community frequently,Web GIS and Cloud GIS applications have drawn attention and have become more efficient for users.In this study,a serverless Cloud GIS framework is implemented for the land valuation platform.In order to store,analyse,and share geospatial data,the Aurora Serverless PostgreSQL database is created on Amazon Web Services(AWS).While adopting Aurora Serverless PostgreSQL as a database management system,a simple point in polygon analysis conducted to compare the performances with Amazon Relational Database Service(RDS)instance.Results showed that the serverless database responded to the query faster and scaled up during high workload to decrease latency.Hence,parcel vector data,which conveys ownership information and land values attributes,is shared directly from the PostGIS database as vector tiles.Besides S3 and AWS Lambda services are used for storing and disseminating raster-based land value map tiles.To visualize all shared data and maps through a web browser,open source web mapping library Mapbox GL JS is used.

A joint function warm-up and request routing scheme for performing confident serverless computing

Serverless computing has become increasingly popular recently due to its cost efficiency and flexibility.However,running serverless computing functions in the cloud can incur high end-to-end service latency and operational costs.Running them on edge servers may significantly reduce service latency but is limited by computing power and memory availability.Given the limitations of cloud and edge environments for performing serverless com-puting,this paper proposes a joint function warm-up and request routing scheme to perform serverless computing functions on edge and cloud collaboratively.The key idea of the new scheme is to maximize the hit ratio of server-less computing requests,thereby reducing the cold-start latency that dominates the overall serving latency.This scheme explicitly considers allocating server memory and operation budget for executing concurrent requests during the scheduling.The proposed scheme has been evaluated through extensive simulations.Its effectiveness has been proved by comparison with the upper-bound results.

基于工作流的无服务边缘计算框架设计与实现

边缘计算具有高性能、体积小、计算力超强等优点。但云设计的边缘计算函数链存在调用时延较长的问题,导致大规模部署和存储能力以及云端和设备端协同困难、无缝对接效果较差。为解决上述问题,基于开源无服务平台OpenFaaS设计了一个无服务边缘计算框架(SECP),为边缘侧提供基于无状态函数的计算服务。对计算服务调用过程采用建模分析,采用了工作流模型对为云设计的无服务计算在边缘侧存在的函数链式调用时延放大的问题改进。性能测试结果表明,SECP在边缘计算场景下有良好的性能表现,降低了边缘计算平台中服务请求的响应时间,解决了无服务计算函数功能单一问题,提高了边缘侧资源的利用率,并改善了边缘侧计算服务部署、维护繁杂的问题。

基于无服务器边缘计算下的服务负载调度算法

随着边缘计算的不断发展,其在资源管理配置方面逐渐出现相关问题,无服务器计算作为一种新的方式可以有效解决边缘计算的相关问题。然而,无服务器计算不具备在分布式边缘场景中高效处理请求所需服务负载调度的能力,针对这一问题,提出了一种基于无服务器边缘计算的服务负载调度算法(service load scheduling algorithm,SLSA)。SLSA的核心是通过隐式建模充分考虑了动态变化的节点状态、负载调度器放置等影响因素来优化整体时延,然后通过改进的平滑加权轮询调度(smooth weighted round robin,SWRR)算法进行服务调度。经仿真实验分析,SLSA在资源消耗上有着明显下降,同时在单城市场景与多城市场景下均有良好的性能表现,其中在单城市场景中相对于集中式轮询调度(round robin centralized,RRC)算法提升了43.01%,在多城市场景中提升了53.81%。实验结果表明,SLSA可以有效降低资源消耗率并提升性能。