Haoran Qiu, Saurabh Jha, Subho Sankar Banerjee, Archit Patke, Chen Wang, H. Franke, Z. Kalbarczyk, R. Iyer
Function-as-a-Service (FaaS) is becoming an increasingly popular cloud-deployment paradigm for serverless computing that frees application developers from managing the infrastructure. At the same time, it allows cloud providers to assert control in workload consolidation, i.e., co-locating multiple containers on the same server, thereby achieving higher server utilization, often at the cost of higher end-to-end function request latency. Interestingly, a key aspect of serverless latency management has not been well studied: the trade-off between application developers' latency goals and the FaaS providers' utilization goals. This paper presents a multi-faceted, measurement-driven study of latency variation in serverless platforms that elucidates this trade-off space. We obtained production measurements by executing FaaS benchmarks on IBM Cloud and a private cloud to study the impact of workload consolidation, queuing delay, and cold starts on the end-to-end function request latency. We draw several conclusions from the characterization results. For example, increasing a container's allocated memory limit from 128 MB to 256 MB reduces the tail latency by 2× but has 1.75× higher power consumption and 59% lower CPU utilization.
{"title":"Is Function-as-a-Service a Good Fit for Latency-Critical Services?","authors":"Haoran Qiu, Saurabh Jha, Subho Sankar Banerjee, Archit Patke, Chen Wang, H. Franke, Z. Kalbarczyk, R. Iyer","doi":"10.1145/3493651.3493666","DOIUrl":"https://doi.org/10.1145/3493651.3493666","url":null,"abstract":"Function-as-a-Service (FaaS) is becoming an increasingly popular cloud-deployment paradigm for serverless computing that frees application developers from managing the infrastructure. At the same time, it allows cloud providers to assert control in workload consolidation, i.e., co-locating multiple containers on the same server, thereby achieving higher server utilization, often at the cost of higher end-to-end function request latency. Interestingly, a key aspect of serverless latency management has not been well studied: the trade-off between application developers' latency goals and the FaaS providers' utilization goals. This paper presents a multi-faceted, measurement-driven study of latency variation in serverless platforms that elucidates this trade-off space. We obtained production measurements by executing FaaS benchmarks on IBM Cloud and a private cloud to study the impact of workload consolidation, queuing delay, and cold starts on the end-to-end function request latency. We draw several conclusions from the characterization results. For example, increasing a container's allocated memory limit from 128 MB to 256 MB reduces the tail latency by 2× but has 1.75× higher power consumption and 59% lower CPU utilization.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126327328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Despite its vast potential, a challenge facing serverless computing's wide-scale adoption is the lack of Service Level Agreements (SLAs) for serverless platforms. This challenge is compounded when composition technologies are employed to construct large applications using chains of functions. Due to the dependency of a chain's performance on each function forming it, a single function's sub-optimal performance can result in performance degradations of the entire chain. This paper sheds light on this problem and provides a categorical classification of the factors that impact a serverless function execution performance. We discuss the challenge of serverless chains' SLA and present the results of leveraging FaaS2F, our proposed serverless SLA framework, to define SLAs for fixed-size and variable-size sequential serverless chains. The validation results demonstrate high accuracy in detecting sub-optimal executions exceeding 79%.
{"title":"SLA for Sequential Serverless Chains: A Machine Learning Approach","authors":"Mohamed Elsakhawy, M. Bauer","doi":"10.1145/3493651.3493671","DOIUrl":"https://doi.org/10.1145/3493651.3493671","url":null,"abstract":"Despite its vast potential, a challenge facing serverless computing's wide-scale adoption is the lack of Service Level Agreements (SLAs) for serverless platforms. This challenge is compounded when composition technologies are employed to construct large applications using chains of functions. Due to the dependency of a chain's performance on each function forming it, a single function's sub-optimal performance can result in performance degradations of the entire chain. This paper sheds light on this problem and provides a categorical classification of the factors that impact a serverless function execution performance. We discuss the challenge of serverless chains' SLA and present the results of leveraging FaaS2F, our proposed serverless SLA framework, to define SLAs for fixed-size and variable-size sequential serverless chains. The validation results demonstrate high accuracy in detecting sub-optimal executions exceeding 79%.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122641698","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Function-as-a-Service or FaaS is a popular delivery model of serverless computing where developers upload code to be executed in the cloud as short running stateless functions. Using smaller functions to decompose processing of larger tasks or workflows introduces the question of how to instrument application control flow to orchestrate an overall task or workflow. In this paper, we examine implications of using different methods to orchestrate the control flow of a serverless data processing pipeline composed as a set of independent FaaS functions. We performed experiments on the AWS Lambda FaaS platform and compared how four different patterns of control flow impact the cost and performance of the pipeline. We investigate control flow using client orchestration, microservice controllers, event-based triggers, and state-machines. Overall, we found that asynchronous methods led to lower orchestration costs, and that event-based orchestration incurred a performance penalty.
{"title":"Implications of Alternative Serverless Application Control Flow Methods","authors":"Sterling D. Quinn, R. Cordingly, W. Lloyd","doi":"10.1145/3493651.3493668","DOIUrl":"https://doi.org/10.1145/3493651.3493668","url":null,"abstract":"Function-as-a-Service or FaaS is a popular delivery model of serverless computing where developers upload code to be executed in the cloud as short running stateless functions. Using smaller functions to decompose processing of larger tasks or workflows introduces the question of how to instrument application control flow to orchestrate an overall task or workflow. In this paper, we examine implications of using different methods to orchestrate the control flow of a serverless data processing pipeline composed as a set of independent FaaS functions. We performed experiments on the AWS Lambda FaaS platform and compared how four different patterns of control flow impact the cost and performance of the pipeline. We investigate control flow using client orchestration, microservice controllers, event-based triggers, and state-machines. Overall, we found that asynchronous methods led to lower orchestration costs, and that event-based orchestration incurred a performance penalty.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"295 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133134132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Simplicity in elastically scalable application development is a key concern addressed by the serverless computing paradigm, in particular the code-level Function-as-a-Service (FaaS). Various FaaSification frameworks demonstrated that marking code methods to streamline their offloading as cloud functions offers a simple bridge to software engineering habits. As application complexity increases, more complex runtime patterns with background activities, such as keeping containerised cloud functions warm to ensure the absence of cold starts, usually require giving up on simplicity and instead investing efforts into orchestrating infrastructure. By bringing infrastructure-as-code concepts into the function source via powerful code annotations, typical orchestration patterns can be simplified again. We evaluate this idea and demonstrate its practical feasibility with FaaS Fusion, an annotations library and transpiler framework for JavaScript.
{"title":"Beyond @CloudFunction: Powerful Code Annotations to Capture Serverless Runtime Patterns","authors":"Raffael Klingler, N. Trifunovic, Josef Spillner","doi":"10.1145/3493651.3493669","DOIUrl":"https://doi.org/10.1145/3493651.3493669","url":null,"abstract":"Simplicity in elastically scalable application development is a key concern addressed by the serverless computing paradigm, in particular the code-level Function-as-a-Service (FaaS). Various FaaSification frameworks demonstrated that marking code methods to streamline their offloading as cloud functions offers a simple bridge to software engineering habits. As application complexity increases, more complex runtime patterns with background activities, such as keeping containerised cloud functions warm to ensure the absence of cold starts, usually require giving up on simplicity and instead investing efforts into orchestrating infrastructure. By bringing infrastructure-as-code concepts into the function source via powerful code annotations, typical orchestration patterns can be simplified again. We evaluate this idea and demonstrate its practical feasibility with FaaS Fusion, an annotations library and transpiler framework for JavaScript.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125678642","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Burstable instances have recently been introduced by cloud providers as a cost-efficient alternative to customers that do not require powerful machines for running their workloads. Unlike conventional instances, the CPU capacity of burstable instances is rate limited, but they can be boosted to their full capacity for small periods when needed. Currently, the majority of cloud providers offer this option as a cheaper solution for their clients. However, little research has been done on the practical usage of these CPU-limited instances. In this paper, we present a novel autoscaling solution that uses burstable instances along with regular instances to handle the queueing arising in traffic and flash crowds. We design BIAS Autoscaler, a state-of-the-art framework that leverages burstable and regular instances for cost-efficient autoscaling and evaluate it on the Google Cloud Platform. We apply our framework to a real-world microservice workload, and conduct extensive experimental evaluations using Google Compute Engines. Experimental results show that BIAS Autoscaler can reduce the overall cost up to 25% and increase resource efficiency by 42% while maintaining the same service quality observed when using conventional instances only.
{"title":"BIAS Autoscaler: Leveraging Burstable Instances for Cost-Effective Autoscaling on Cloud Systems","authors":"Jaime Dantas, Hamzeh Khazaei, Marin Litoiu","doi":"10.1145/3493651.3493667","DOIUrl":"https://doi.org/10.1145/3493651.3493667","url":null,"abstract":"Burstable instances have recently been introduced by cloud providers as a cost-efficient alternative to customers that do not require powerful machines for running their workloads. Unlike conventional instances, the CPU capacity of burstable instances is rate limited, but they can be boosted to their full capacity for small periods when needed. Currently, the majority of cloud providers offer this option as a cheaper solution for their clients. However, little research has been done on the practical usage of these CPU-limited instances. In this paper, we present a novel autoscaling solution that uses burstable instances along with regular instances to handle the queueing arising in traffic and flash crowds. We design BIAS Autoscaler, a state-of-the-art framework that leverages burstable and regular instances for cost-efficient autoscaling and evaluate it on the Google Cloud Platform. We apply our framework to a real-world microservice workload, and conduct extensive experimental evaluations using Google Compute Engines. Experimental results show that BIAS Autoscaler can reduce the overall cost up to 25% and increase resource efficiency by 42% while maintaining the same service quality observed when using conventional instances only.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122541605","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Over the past couple of years, serverless computing has become a popular way of structuring and deploying applications in the cloud. However, several practical and research challenges remain. In this paper, we provide the first step to address two open issues. We developed a simple extension language (SFL) and a compiler to enable software developers to write entire serverless applications as one piece. The compiler generates necessary orchestration code that automatically binds several functions together. In addition, the SFL tools allow programmers to write stateful serverless functions with the compiler generating supporting cloud infrastructure for the storage and access of the application state. We evaluate our system using simple benchmark programs, comparing the resulting performance to Azure durable functions, which directly supports statefulness. The execution times we see in our unoptimized code are only slightly worse than what we measure on the Azure platform. Overall execution times are considerably better due to better scheduling by AWS Lambda than the Azure durable functions.
{"title":"SFL: A Compiler for Generating Stateful AWS Lambda Serverless Applications","authors":"Lukas Brand, Markus U. Mock","doi":"10.1145/3493651.3493670","DOIUrl":"https://doi.org/10.1145/3493651.3493670","url":null,"abstract":"Over the past couple of years, serverless computing has become a popular way of structuring and deploying applications in the cloud. However, several practical and research challenges remain. In this paper, we provide the first step to address two open issues. We developed a simple extension language (SFL) and a compiler to enable software developers to write entire serverless applications as one piece. The compiler generates necessary orchestration code that automatically binds several functions together. In addition, the SFL tools allow programmers to write stateful serverless functions with the compiler generating supporting cloud infrastructure for the storage and access of the application state. We evaluate our system using simple benchmark programs, comparing the resulting performance to Azure durable functions, which directly supports statefulness. The execution times we see in our unoptimized code are only slightly worse than what we measure on the Azure platform. Overall execution times are considerably better due to better scheduling by AWS Lambda than the Azure durable functions.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116445766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Serverless computing offers a pay-per-use model with high elasticity and automatic scaling for a wide range of applications. Since cloud providers abstract most of the underlying infrastructure, these services work similarly to black-boxes. As a result, users can influence the resources allocated to their functions, but might not be aware that they have to parallelize them to profit from the additionally allocated virtual CPUs (vCPUs). In this paper, we analyze the impact of parallelization within a single function and container instance for AWS Lambda, Google Cloud Functions (GCF), and Google Cloud Run (GCR). We focus on compute-intensive workloads since they benefit greatly from parallelization. Furthermore, we investigate the correlation between the number of allocated CPU cores and vCPUs in serverless environments. Our results show that the number of available cores to a function/container instance does not always equal the number of allocated vCPUs. By parallelizing serverless workloads, we observed cost savings up to 81% for AWS Lambda, 49% for GCF, and 69.8% for GCR.
无服务器计算为广泛的应用程序提供了具有高弹性和自动扩展的按使用付费模型。由于云提供商抽象了大部分底层基础设施，因此这些服务的工作方式类似于黑盒。因此，用户可以影响分配给其函数的资源，但可能不知道他们必须并行化这些资源才能从额外分配的虚拟cpu (vcpu)中获利。在本文中，我们分析了AWS Lambda、Google Cloud Functions (GCF)和Google Cloud Run (GCR)在单个函数和容器实例中并行化的影响。我们主要关注计算密集型工作负载，因为它们从并行化中受益匪浅。此外，我们还研究了在无服务器环境中分配的CPU内核数量与vcpu之间的相关性。我们的结果表明，函数/容器实例的可用内核数量并不总是等于分配的vcpu数量。通过并行化无服务器工作负载，我们发现AWS Lambda节省了81%的成本，GCF节省了49%，GCR节省了69.8%。
{"title":"Towards Demystifying Intra-Function Parallelism in Serverless Computing","authors":"M. Kiener, Mohak Chadha, M. Gerndt","doi":"10.1145/3493651.3493672","DOIUrl":"https://doi.org/10.1145/3493651.3493672","url":null,"abstract":"Serverless computing offers a pay-per-use model with high elasticity and automatic scaling for a wide range of applications. Since cloud providers abstract most of the underlying infrastructure, these services work similarly to black-boxes. As a result, users can influence the resources allocated to their functions, but might not be aware that they have to parallelize them to profit from the additionally allocated virtual CPUs (vCPUs). In this paper, we analyze the impact of parallelization within a single function and container instance for AWS Lambda, Google Cloud Functions (GCF), and Google Cloud Run (GCR). We focus on compute-intensive workloads since they benefit greatly from parallelization. Furthermore, we investigate the correlation between the number of allocated CPU cores and vCPUs in serverless environments. Our results show that the number of available cores to a function/container instance does not always equal the number of allocated vCPUs. By parallelizing serverless workloads, we observed cost savings up to 81% for AWS Lambda, 49% for GCF, and 69.8% for GCR.","PeriodicalId":270470,"journal":{"name":"Proceedings of the Seventh International Workshop on Serverless Computing (WoSC7) 2021","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117091280","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：