Sebastian Maier, Timo Hönig, Peter Wägemann, Wolfgang Schröder-Preikschat
Today's systems offer an increasing number of processor cores, however, the chance to operate them efficiently by dedicating cores to specific tasks is often missed. Instead, mixed workloads are processed by each core which leads to system noise (i.e., interferences, scheduling overheads) and yields subpar performance, only. We therefore propose a system design based on Asynchronous Abstract Machines (AAMs). An AAM features a light-weight scheduler and is dedicated to a specific group of tasks with common characteristics (i.e., shared code and data). It offers an asynchronous, task-based interface for efficient interaction between AAMs. Just like applications are built from AAMs, even the OS itself consists of AAMs that are interfaced by applications via asynchronous messages instead of synchronous system calls. A dedicated OS component, which is aware of all AAMs in the system, is responsible for dynamic and exclusive allocation of cores to AAMs depending on their current workload. Thus, cores rarely switch between heterogeneous workloads of different AAMs. And, at the same time, frequent switches between homogeneous tasks become fast, local operations of an AAM, which do not involve the OS kernel. In this paper, we describe shortcomings of existing operating systems, our new system design concept, and present evaluation results of our prototype implementation.
{"title":"Asynchronous Abstract Machines: Anti-noise System Software for Many-core Processors","authors":"Sebastian Maier, Timo Hönig, Peter Wägemann, Wolfgang Schröder-Preikschat","doi":"10.1145/3322789.3328744","DOIUrl":"https://doi.org/10.1145/3322789.3328744","url":null,"abstract":"Today's systems offer an increasing number of processor cores, however, the chance to operate them efficiently by dedicating cores to specific tasks is often missed. Instead, mixed workloads are processed by each core which leads to system noise (i.e., interferences, scheduling overheads) and yields subpar performance, only. We therefore propose a system design based on Asynchronous Abstract Machines (AAMs). An AAM features a light-weight scheduler and is dedicated to a specific group of tasks with common characteristics (i.e., shared code and data). It offers an asynchronous, task-based interface for efficient interaction between AAMs. Just like applications are built from AAMs, even the OS itself consists of AAMs that are interfaced by applications via asynchronous messages instead of synchronous system calls. A dedicated OS component, which is aware of all AAMs in the system, is responsible for dynamic and exclusive allocation of cores to AAMs depending on their current workload. Thus, cores rarely switch between heterogeneous workloads of different AAMs. And, at the same time, frequent switches between homogeneous tasks become fast, local operations of an AAM, which do not involve the OS kernel. In this paper, we describe shortcomings of existing operating systems, our new system design concept, and present evaluation results of our prototype implementation.","PeriodicalId":365438,"journal":{"name":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126057180","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}
Building efficient and scalable system software, especially performance analysis and monitoring, for large-scale systems, is increasingly important both for the developers of parallel applications and the designers of next-generation HPC systems. However, conventional performance tools suffer from significant time/space overhead due to the ever-increasing problem size and system scale. For instance, memory monitoring is of critical use in understanding applications and evaluating systems. Due to the dynamic nature in programs' memory accesses, common practice today leaves large amounts of address examination and data recording at runtime, at the cost of substantial performance overhead. On the other hand, the cost of source code analysis is independent of the problem size and system scale, making it very appealing for large-scale performance analysis. Inspired by this observation, we have designed a series of light-weight system software for HPC systems, such as a memory access monitoring tool, a performance variance detection tool , and a communication trace compression tool. In this talk, I will share our experience on building these tools through combining static analysis and runtime analysis and also point out main challenges in this direction.
{"title":"HPC System Software Enhanced by Source Code Analysis","authors":"Jidong Zhai","doi":"10.1145/3322789.3328741","DOIUrl":"https://doi.org/10.1145/3322789.3328741","url":null,"abstract":"Building efficient and scalable system software, especially performance analysis and monitoring, for large-scale systems, is increasingly important both for the developers of parallel applications and the designers of next-generation HPC systems. However, conventional performance tools suffer from significant time/space overhead due to the ever-increasing problem size and system scale. For instance, memory monitoring is of critical use in understanding applications and evaluating systems. Due to the dynamic nature in programs' memory accesses, common practice today leaves large amounts of address examination and data recording at runtime, at the cost of substantial performance overhead. On the other hand, the cost of source code analysis is independent of the problem size and system scale, making it very appealing for large-scale performance analysis. Inspired by this observation, we have designed a series of light-weight system software for HPC systems, such as a memory access monitoring tool, a performance variance detection tool , and a communication trace compression tool. In this talk, I will share our experience on building these tools through combining static analysis and runtime analysis and also point out main challenges in this direction.","PeriodicalId":365438,"journal":{"name":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126107855","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}
Boyang Li, Sudheer Chunduri, K. Harms, Yuping Fan, Z. Lan
Application performance variability caused by network contention is a major issue on dragonfly based systems. This work-in-progress study makes two contributions. First, we analyze real workload logs and conduct application experiments on the production system Theta at Argonne to evaluate application performance variability. We find a strong correlation between system utilization and performance variability where a high system utilization (e.g., above 95%) can cause up to 21% degradation in application performance. Next, driven by this key finding, we investigate a scheduling policy to mitigate workload interference by leveraging the fact that production systems often exhibit diurnal utilization behavior and not all users are in a hurry for job completion. Preliminary results show that this scheduling design is capable of improving system productivity (measured by scheduling makespan) as well as improving user-level scheduling metrics such as user wait time and job slowdown.
{"title":"The Effect of System Utilization on Application Performance Variability","authors":"Boyang Li, Sudheer Chunduri, K. Harms, Yuping Fan, Z. Lan","doi":"10.1145/3322789.3328743","DOIUrl":"https://doi.org/10.1145/3322789.3328743","url":null,"abstract":"Application performance variability caused by network contention is a major issue on dragonfly based systems. This work-in-progress study makes two contributions. First, we analyze real workload logs and conduct application experiments on the production system Theta at Argonne to evaluate application performance variability. We find a strong correlation between system utilization and performance variability where a high system utilization (e.g., above 95%) can cause up to 21% degradation in application performance. Next, driven by this key finding, we investigate a scheduling policy to mitigate workload interference by leveraging the fact that production systems often exhibit diurnal utilization behavior and not all users are in a hurry for job completion. Preliminary results show that this scheduling design is capable of improving system productivity (measured by scheduling makespan) as well as improving user-level scheduling metrics such as user wait time and job slowdown.","PeriodicalId":365438,"journal":{"name":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130852493","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}
Specialized operating systems have enjoyed a recent revival driven both by a pressing need to rethink the system software stack in several domains and by the convenience and flexibility that on-demand infrastructure and virtual execution environments offer. Several barriers exist which curtail the widespread adoption of such highly specialized systems, but perhaps the most consequential of them is that these systems are simply difficult to use. In this paper we discuss the challenges faced by specialized OSes, both for HPC and more broadly, and argue that what is needed to make them practically useful is a reasonable development and deployment model that will form the foundation for a kernel ecosystem that allows intrepid developers to discover, experiment with, contribute to, and write programs for available kernel frameworks while safely ignoring complexities such as provisioning, deployment, cross-compilation, and interface compatibility. We argue that such an ecosystem would allow more developers of highly tuned applications to reap the performance benefits of specialized kernels.
{"title":"Towards a Practical Ecosystem of Specialized OS Kernels","authors":"Conghao Liu, Kyle C. Hale","doi":"10.1145/3322789.3328742","DOIUrl":"https://doi.org/10.1145/3322789.3328742","url":null,"abstract":"Specialized operating systems have enjoyed a recent revival driven both by a pressing need to rethink the system software stack in several domains and by the convenience and flexibility that on-demand infrastructure and virtual execution environments offer. Several barriers exist which curtail the widespread adoption of such highly specialized systems, but perhaps the most consequential of them is that these systems are simply difficult to use. In this paper we discuss the challenges faced by specialized OSes, both for HPC and more broadly, and argue that what is needed to make them practically useful is a reasonable development and deployment model that will form the foundation for a kernel ecosystem that allows intrepid developers to discover, experiment with, contribute to, and write programs for available kernel frameworks while safely ignoring complexities such as provisioning, deployment, cross-compilation, and interface compatibility. We argue that such an ecosystem would allow more developers of highly tuned applications to reap the performance benefits of specialized kernels.","PeriodicalId":365438,"journal":{"name":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","volume":"112 ","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114001178","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":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","authors":"","doi":"10.1145/3322789","DOIUrl":"https://doi.org/10.1145/3322789","url":null,"abstract":"","PeriodicalId":365438,"journal":{"name":"Proceedings of the 9th International Workshop on Runtime and Operating Systems for Supercomputers","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115631808","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：