Matt Baughman, Rohan Kumar, Ian T Foster, K. Chard
Recent advances in networking technology and serverless architectures have enabled automated distribution of compute workloads at the function level. As heterogeneity and physical distribution of computing resources increase, so too does the need to effectively use those resources. This is especially true when leveraging multiple compute resources in the form of local, distributed, and cloud resources. Adding to the complexity of the problem is different notions of "cost" when it comes to using these resources. Tradeoffs exist due to the inherent difference between costs of computation for the end user. For example, deploying a workload on the cloud could be much faster than using local resources but using the cloud incurs a financial cost. Here, the end user is presented with the tradeoff between time and money. We describe preliminary work towards Delts+, a framework that integrates multidimensional cost objectives, cost tradeoffs, and optimization under constraints.
{"title":"Expanding Cost-Aware Function Execution with Multidimensional Notions of Cost","authors":"Matt Baughman, Rohan Kumar, Ian T Foster, K. Chard","doi":"10.1145/3452413.3464790","DOIUrl":"https://doi.org/10.1145/3452413.3464790","url":null,"abstract":"Recent advances in networking technology and serverless architectures have enabled automated distribution of compute workloads at the function level. As heterogeneity and physical distribution of computing resources increase, so too does the need to effectively use those resources. This is especially true when leveraging multiple compute resources in the form of local, distributed, and cloud resources. Adding to the complexity of the problem is different notions of \"cost\" when it comes to using these resources. Tradeoffs exist due to the inherent difference between costs of computation for the end user. For example, deploying a workload on the cloud could be much faster than using local resources but using the cloud incurs a financial cost. Here, the end user is presented with the tradeoff between time and money. We describe preliminary work towards Delts+, a framework that integrates multidimensional cost objectives, cost tradeoffs, and optimization under constraints.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115064680","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}
Jacek Kusnierz, M. Malawski, V. Padulano, E. T. Saavedra, P. Alonso-Jordá
The High-Energy Physics experiments at CERN produce a high volume of data. It is not possible to analyze big chunks of it within a reasonable time by any single machine. The ROOT framework was recently extended with the distributed computing capabilities for massively parallelized RDataFrame applications. This approach, using the MapReduce pattern underneath, made the heavy computations much more approachable even for the newcomers. This paper explores the possibility of running such analyses on serverless services in public cloud using a purely stateless environment. So far, the distributed approaches used by RDataFrame relied on stateful, fully managed computing frameworks like Apache Spark. Here we show that our newly developed tool is able to use perfectly stateless cloud functions, demonstrating the excellent speedup in parallel stage of processing in our benchmarks.
{"title":"Distributed Parallel Analysis Engine for High Energy Physics Using AWS Lambda","authors":"Jacek Kusnierz, M. Malawski, V. Padulano, E. T. Saavedra, P. Alonso-Jordá","doi":"10.1145/3452413.3464788","DOIUrl":"https://doi.org/10.1145/3452413.3464788","url":null,"abstract":"The High-Energy Physics experiments at CERN produce a high volume of data. It is not possible to analyze big chunks of it within a reasonable time by any single machine. The ROOT framework was recently extended with the distributed computing capabilities for massively parallelized RDataFrame applications. This approach, using the MapReduce pattern underneath, made the heavy computations much more approachable even for the newcomers. This paper explores the possibility of running such analyses on serverless services in public cloud using a purely stateless environment. So far, the distributed approaches used by RDataFrame relied on stateful, fully managed computing frameworks like Apache Spark. Here we show that our newly developed tool is able to use perfectly stateless cloud functions, demonstrating the excellent speedup in parallel stage of processing in our benchmarks.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130500906","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}
K. Satzke, I. E. Akkus, Ruichuan Chen, Ivica Rimac, M. Stein, Andre Beck, Paarijaat Aditya, M. Vanga, V. Hilt
Serverless computing has emerged as a new cloud computing paradigm, where an application consists of individual functions that can be separately managed and executed. However, the function development environment of all serverless computing frameworks at present is CPU-based. In this paper, we propose to extend the open-sourced KNIX high-performance serverless framework so that it can execute functions on shared GPU cluster resources. We have evaluated the performance impacts on the extended KNIX system by measuring overheads and penalties incurred using different deep learning frameworks.
{"title":"Efficient GPU Sharing for Serverless Workflows","authors":"K. Satzke, I. E. Akkus, Ruichuan Chen, Ivica Rimac, M. Stein, Andre Beck, Paarijaat Aditya, M. Vanga, V. Hilt","doi":"10.1145/3452413.3464785","DOIUrl":"https://doi.org/10.1145/3452413.3464785","url":null,"abstract":"Serverless computing has emerged as a new cloud computing paradigm, where an application consists of individual functions that can be separately managed and executed. However, the function development environment of all serverless computing frameworks at present is CPU-based. In this paper, we propose to extend the open-sourced KNIX high-performance serverless framework so that it can execute functions on shared GPU cluster resources. We have evaluated the performance impacts on the extended KNIX system by measuring overheads and penalties incurred using different deep learning frameworks.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132646548","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}
This paper provides a first presentation of Apollo, a runtime system for serverless function compositions distributed across the cloud-edge-IoT continuum. Apollo's modular design enables a fine-grained decomposition of the runtime implementation(scheduling, data transmission, etc.) of the application, so that each of the numerous implementation decisions can be optimized separately, fully exploiting the potential for the optimization of the overall performance and costs. Apollo features (a) a flexible model of the application and the available resources and (b) an implementation process based on a large set of independent agents. This flexible structure enables distributing not only the processing, but the implementation process itself across a large number of resources, each running an independent Apollo instance. The ability to flexibly determine the placement of implementation actions opens up new optimization opportunities, while at the same time providing access to greater computing power for optimizing challenging decisions such as task scheduling and the placement and routing of data.
{"title":"Apollo: Modular and Distributed Runtime System for Serverless Function Compositions on Cloud, Edge, and IoT Resources","authors":"Fedor Smirnov, Behnaz Pourmohseni, T. Fahringer","doi":"10.1145/3452413.3464793","DOIUrl":"https://doi.org/10.1145/3452413.3464793","url":null,"abstract":"This paper provides a first presentation of Apollo, a runtime system for serverless function compositions distributed across the cloud-edge-IoT continuum. Apollo's modular design enables a fine-grained decomposition of the runtime implementation(scheduling, data transmission, etc.) of the application, so that each of the numerous implementation decisions can be optimized separately, fully exploiting the potential for the optimization of the overall performance and costs. Apollo features (a) a flexible model of the application and the available resources and (b) an implementation process based on a large set of independent agents. This flexible structure enables distributing not only the processing, but the implementation process itself across a large number of resources, each running an independent Apollo instance. The ability to flexibly determine the placement of implementation actions opens up new optimization opportunities, while at the same time providing access to greater computing power for optimizing challenging decisions such as task scheduling and the placement and routing of data.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133739822","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}
Dheeraj Chahal, S. Palepu, Mayank Mishra, Rekha Singhal
Cloud services have an auto-scaling feature for load balancing to meet the performance requirements of an application. Existing auto-scaling techniques are based on upscaling and downscaling cloud resources to distribute the dynamically varying workloads. However, bursty workloads pose many challenges for auto-scaling and sometimes result in Service Level Agreement (SLA) violations. Furthermore, over-provisioning or under-provisioning cloud resources to address dynamically evolving workloads results in performance degradation and cost escalation. In this work, we present a workload characterization based approach for scheduling the bursty workload on a highly scalable serverless architecture in conjunction with a machine learning (ML) platform. We present the use of Amazon Web Services (AWS) ML platform SageMaker and serverless computing platform Lambda for load balancing the inference workload to avoid SLA violations. We evaluate our approach using a recommender system that is based on a deep learning model for inference.
云服务具有用于负载平衡的自动扩展特性，以满足应用程序的性能需求。现有的自动扩展技术是基于升级和缩减云资源来分配动态变化的工作负载。然而，突发工作负载给自动扩展带来了许多挑战，有时会导致违反服务水平协议(SLA)。此外，为解决动态发展的工作负载而过度配置或配置不足的云资源会导致性能下降和成本上升。在这项工作中，我们提出了一种基于工作负载特征的方法，用于与机器学习(ML)平台一起在高度可扩展的无服务器架构上调度突发工作负载。我们介绍了使用Amazon Web Services (AWS) ML平台SageMaker和无服务器计算平台Lambda来负载平衡推理工作负载，以避免违反SLA。我们使用基于深度学习推理模型的推荐系统来评估我们的方法。
{"title":"SLA-aware Workload Scheduling Using Hybrid Cloud Services","authors":"Dheeraj Chahal, S. Palepu, Mayank Mishra, Rekha Singhal","doi":"10.1145/3452413.3464789","DOIUrl":"https://doi.org/10.1145/3452413.3464789","url":null,"abstract":"Cloud services have an auto-scaling feature for load balancing to meet the performance requirements of an application. Existing auto-scaling techniques are based on upscaling and downscaling cloud resources to distribute the dynamically varying workloads. However, bursty workloads pose many challenges for auto-scaling and sometimes result in Service Level Agreement (SLA) violations. Furthermore, over-provisioning or under-provisioning cloud resources to address dynamically evolving workloads results in performance degradation and cost escalation. In this work, we present a workload characterization based approach for scheduling the bursty workload on a highly scalable serverless architecture in conjunction with a machine learning (ML) platform. We present the use of Amazon Web Services (AWS) ML platform SageMaker and serverless computing platform Lambda for load balancing the inference workload to avoid SLA violations. We evaluate our approach using a recommender system that is based on a deep learning model for inference.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130229031","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}
{"title":"Session details: Session: Full Papers","authors":"K. Chard","doi":"10.1145/3470764","DOIUrl":"https://doi.org/10.1145/3470764","url":null,"abstract":"","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"161 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124512793","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}
The historical motivation for serverless comes from internet-of-things, smartphone client server, and the objective of simplifying programming (no provisioning) and scale-down (pay-for-use). These applications are generally low-performance best-effort. However, the serverless model enables flexible software architectures suitable for a wide range of applications that demand high-performance and guaranteed performance. We have studied three such applications - scientific data streaming, virtual/augmented reality, and document annotation. We describe how each can be cast in a serverless software architecture and how the application performance requirements translate into high performance requirements (invocation rate, low and predictable latency) for the underlying serverless system implementation. These applications can require invocations rates as high as millions per second (40 MHz) and latency deadlines below a microsecond (300 ns), and furthermore require performance predictability. All of these capabilities are far in excess of today's commercial serverless offerings and represent interesting research challenges.
{"title":"Motivating High Performance Serverless Workloads","authors":"H. Nguyen, Zhifei Yang, A. Chien","doi":"10.1145/3452413.3464786","DOIUrl":"https://doi.org/10.1145/3452413.3464786","url":null,"abstract":"The historical motivation for serverless comes from internet-of-things, smartphone client server, and the objective of simplifying programming (no provisioning) and scale-down (pay-for-use). These applications are generally low-performance best-effort. However, the serverless model enables flexible software architectures suitable for a wide range of applications that demand high-performance and guaranteed performance. We have studied three such applications - scientific data streaming, virtual/augmented reality, and document annotation. We describe how each can be cast in a serverless software architecture and how the application performance requirements translate into high performance requirements (invocation rate, low and predictable latency) for the underlying serverless system implementation. These applications can require invocations rates as high as millions per second (40 MHz) and latency deadlines below a microsecond (300 ns), and furthermore require performance predictability. All of these capabilities are far in excess of today's commercial serverless offerings and represent interesting research challenges.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127041491","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}
S. Yao, Muhammad Ali Gulzar, Liqing Zhang, A. Butt
Function-as-a-Service (FaaS) and the serverless computing model offer a powerful abstraction for supporting large-scale applications in the cloud. A major hurdle in this context is that it is non-trivial to transform an application, even an already containerized one, to a FaaS implementation. In this paper, we take the first step towards supporting easier and efficient application transformation to FaaS. We present a systematic scheme to transform applications written in Python into a set of functions that can then be automatically deployed atop platforms such as AWS Lamda. We target a Bioinformatics cyberinfrastructure pipeline, CIWARS, that provides waste-water analysis for the identification of antibiotic-resistant bacteria and viruses such as SARS-CoV-2. Based on our experience with enabling FaaS-based CIWARS, we develop a methodology that would help the conversion of other similar applications to the FaaS model. Our evaluation shows that our approach can correctly transform CIWARS to FaaS, and the new FaaS-based CIWARS incurs only negligible(≤2%) less than 2% overhead for representative workloads.
{"title":"Towards a Serverless Bioinformatics Cyberinfrastructure Pipeline","authors":"S. Yao, Muhammad Ali Gulzar, Liqing Zhang, A. Butt","doi":"10.1145/3452413.3464787","DOIUrl":"https://doi.org/10.1145/3452413.3464787","url":null,"abstract":"Function-as-a-Service (FaaS) and the serverless computing model offer a powerful abstraction for supporting large-scale applications in the cloud. A major hurdle in this context is that it is non-trivial to transform an application, even an already containerized one, to a FaaS implementation. In this paper, we take the first step towards supporting easier and efficient application transformation to FaaS. We present a systematic scheme to transform applications written in Python into a set of functions that can then be automatically deployed atop platforms such as AWS Lamda. We target a Bioinformatics cyberinfrastructure pipeline, CIWARS, that provides waste-water analysis for the identification of antibiotic-resistant bacteria and viruses such as SARS-CoV-2. Based on our experience with enabling FaaS-based CIWARS, we develop a methodology that would help the conversion of other similar applications to the FaaS model. Our evaluation shows that our approach can correctly transform CIWARS to FaaS, and the new FaaS-based CIWARS incurs only negligible(≤2%) less than 2% overhead for representative workloads.","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"273 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133505565","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}
{"title":"Session details: Session: Short Papers","authors":"Zhuozhao Li","doi":"10.1145/3470763","DOIUrl":"https://doi.org/10.1145/3470763","url":null,"abstract":"","PeriodicalId":339058,"journal":{"name":"Proceedings of the 1st Workshop on High Performance Serverless Computing","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114633540","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：