Yaochen Hu, Yushi Wang, Bang Liu, Di Niu, Cheng Huang
Erasure coding has been used in storage systems to enhance data durability at a lower storage overhead. However, these systems suffer from long access latency tails due to a lack of flexible load balancing mechanisms and passively launched degraded reads when the original storage node of the requested data becomes a hotspot. We provide a new perspective to load balancing in coded storage systems by proactively and intelligently launching degraded reads and propose a variety of schemes to make optimal decisions either per request or across requests statistically. Experiments on a 98-machine cluster based on the request traces of 12 million objects collected from Windows Azure Storage (WAS) show that our schemes can reduce the median latency by 44.7% and the 95th-percentile tail latency by 77.8% in coded storage systems.
{"title":"Latency reduction and load balancing in coded storage systems","authors":"Yaochen Hu, Yushi Wang, Bang Liu, Di Niu, Cheng Huang","doi":"10.1145/3127479.3131623","DOIUrl":"https://doi.org/10.1145/3127479.3131623","url":null,"abstract":"Erasure coding has been used in storage systems to enhance data durability at a lower storage overhead. However, these systems suffer from long access latency tails due to a lack of flexible load balancing mechanisms and passively launched degraded reads when the original storage node of the requested data becomes a hotspot. We provide a new perspective to load balancing in coded storage systems by proactively and intelligently launching degraded reads and propose a variety of schemes to make optimal decisions either per request or across requests statistically. Experiments on a 98-machine cluster based on the request traces of 12 million objects collected from Windows Azure Storage (WAS) show that our schemes can reduce the median latency by 44.7% and the 95th-percentile tail latency by 77.8% in coded storage systems.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"13 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78057001","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. Costache, Tommaso Madonia, A. Tantawi, M. Steinder
Hosting data analytics services is challenging as their workload is often composed of on-line (e.g., interactive or streaming), requiring fast on-demand provisioning, and batch jobs. As workload demand fluctuations lead to varying idle capacity, efficient resource management is difficult, in particular given different provider objectives, e.g., utilization, revenue.
{"title":"Batch spot market for data analytics cloud providers","authors":"S. Costache, Tommaso Madonia, A. Tantawi, M. Steinder","doi":"10.1145/3127479.3134348","DOIUrl":"https://doi.org/10.1145/3127479.3134348","url":null,"abstract":"Hosting data analytics services is challenging as their workload is often composed of on-line (e.g., interactive or streaming), requiring fast on-demand provisioning, and batch jobs. As workload demand fluctuations lead to varying idle capacity, efficient resource management is difficult, in particular given different provider objectives, e.g., utilization, revenue.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77194527","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}
Yonatan Gottesman, Aran Bergman, Edward Bortnikov, Eshcar Hillel, I. Keidar, Ohad Shacham
As transaction processing services begin to be used in new application domains, low transaction latency becomes an important consideration. Motivated by such use cases we developed Fragola, a highly scalable low-latency and high-throughput transaction processing engine for Apache HBase. Similarly to other modern transaction managers, Fragola provides a variant of generalized snapshot isolation (SI), which scales better than traditional serializability implementations.
{"title":"Fragola: low-latency transactions in distributed data stores","authors":"Yonatan Gottesman, Aran Bergman, Edward Bortnikov, Eshcar Hillel, I. Keidar, Ohad Shacham","doi":"10.1145/3127479.3132686","DOIUrl":"https://doi.org/10.1145/3127479.3132686","url":null,"abstract":"As transaction processing services begin to be used in new application domains, low transaction latency becomes an important consideration. Motivated by such use cases we developed Fragola, a highly scalable low-latency and high-throughput transaction processing engine for Apache HBase. Similarly to other modern transaction managers, Fragola provides a variant of generalized snapshot isolation (SI), which scales better than traditional serializability implementations.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84137971","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}
Cloud computing needs to process and analyze massive high-dimensional data in a real-time manner. Approximate queries in cloud computing systems can provide timely queried results with acceptable accuracy, thus alleviating the consumption of a large amount of resources. Locality Sensitive Hashing (LSH) is able to maintain the data locality and support approximate queries. However, due to randomly choosing hash functions, LSH has to use too many functions to guarantee the query accuracy. The extra computation and storage overheads exacerbate the real performance of LSH. In order to reduce the overheads and deliver high performance, we propose a distribution-aware scheme, called DLSH, to offer cost-effective approximate nearest neighbor query service for cloud computing. The idea of DLSH is to leverage the principal components of the data distribution as the projection vectors of hash functions in LSH, further quantify the weight of each hash function and adjust the interval value in each hash table. We then refine the queried result set based on the hit frequency to significantly decrease the time overhead of distance computation. Extensive experiments in a large-scale cloud computing testbed demonstrate significant improvements in terms of multiple system performance metrics. We have released the source code of DLSH for public use.
{"title":"DLSH: a distribution-aware LSH scheme for approximate nearest neighbor query in cloud computing","authors":"Yuanyuan Sun, Yu Hua, Xue Liu, Shunde Cao, Pengfei Zuo","doi":"10.1145/3127479.3127485","DOIUrl":"https://doi.org/10.1145/3127479.3127485","url":null,"abstract":"Cloud computing needs to process and analyze massive high-dimensional data in a real-time manner. Approximate queries in cloud computing systems can provide timely queried results with acceptable accuracy, thus alleviating the consumption of a large amount of resources. Locality Sensitive Hashing (LSH) is able to maintain the data locality and support approximate queries. However, due to randomly choosing hash functions, LSH has to use too many functions to guarantee the query accuracy. The extra computation and storage overheads exacerbate the real performance of LSH. In order to reduce the overheads and deliver high performance, we propose a distribution-aware scheme, called DLSH, to offer cost-effective approximate nearest neighbor query service for cloud computing. The idea of DLSH is to leverage the principal components of the data distribution as the projection vectors of hash functions in LSH, further quantify the weight of each hash function and adjust the interval value in each hash table. We then refine the queried result set based on the hit frequency to significantly decrease the time overhead of distance computation. Extensive experiments in a large-scale cloud computing testbed demonstrate significant improvements in terms of multiple system performance metrics. We have released the source code of DLSH for public use.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"26 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84627734","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}
Yutaro Hayakawa, L. Eggert, Michio Honda, Douglas J. Santry
In datacenters, workload throughput is often constrained by the attachment bandwidth of proxy servers, despite the much higher aggregate bandwidth of backend servers. We introduce a novel architecture that addresses this problem by combining programmable network switches with a controller that together act as a network "Prism" that can transparently redirect individual client transactions to different backend servers. Unlike traditional proxy approaches, with Prism, transaction payload data is exchanged directly between clients and backend servers, which eliminates the proxy bottleneck. Because the controller only handles transactional metadata, it should scale to much higher transaction rates than traditional proxies. An experimental evaluation with a prototype implementation demonstrates correctness of operation, improved bandwidth utilization and low packet transformation overheads even in software.
{"title":"Prism: a proxy architecture for datacenter networks","authors":"Yutaro Hayakawa, L. Eggert, Michio Honda, Douglas J. Santry","doi":"10.1145/3127479.3127480","DOIUrl":"https://doi.org/10.1145/3127479.3127480","url":null,"abstract":"In datacenters, workload throughput is often constrained by the attachment bandwidth of proxy servers, despite the much higher aggregate bandwidth of backend servers. We introduce a novel architecture that addresses this problem by combining programmable network switches with a controller that together act as a network \"Prism\" that can transparently redirect individual client transactions to different backend servers. Unlike traditional proxy approaches, with Prism, transaction payload data is exchanged directly between clients and backend servers, which eliminates the proxy bottleneck. Because the controller only handles transactional metadata, it should scale to much higher transaction rates than traditional proxies. An experimental evaluation with a prototype implementation demonstrates correctness of operation, improved bandwidth utilization and low packet transformation overheads even in software.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"23 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89747825","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}
In a cloud data center, a single physical machine simultaneously executes dozens of highly heterogeneous tasks. Such colocation results in more efficient utilization of machines, but, when tasks' requirements exceed available resources, some of the tasks might be throttled down or preempted. We analyze version 2.1 of the Google cluster trace that shows short-term (1 second) task CPU usage. Contrary to the assumptions taken by many theoretical studies, we demonstrate that the empirical distributions do not follow any single distribution. However, high percentiles of the total processor usage (summed over at least 10 tasks) can be reasonably estimated by the Gaussian distribution. We use this result for a probabilistic fit test, called the Gaussian Percentile Approximation (GPA), for standard bin-packing algorithms. To check whether a new task will fit into a machine, GPA checks whether the resulting distribution's percentile corresponding to the requested service level objective, SLO is still below the machine's capacity. In our simulation experiments, GPA resulted in colocations exceeding the machines' capacity with a frequency similar to the requested SLO.
{"title":"SLO-aware colocation of data center tasks based on instantaneous processor requirements","authors":"P. Janus, K. Rządca","doi":"10.1145/3127479.3132244","DOIUrl":"https://doi.org/10.1145/3127479.3132244","url":null,"abstract":"In a cloud data center, a single physical machine simultaneously executes dozens of highly heterogeneous tasks. Such colocation results in more efficient utilization of machines, but, when tasks' requirements exceed available resources, some of the tasks might be throttled down or preempted. We analyze version 2.1 of the Google cluster trace that shows short-term (1 second) task CPU usage. Contrary to the assumptions taken by many theoretical studies, we demonstrate that the empirical distributions do not follow any single distribution. However, high percentiles of the total processor usage (summed over at least 10 tasks) can be reasonably estimated by the Gaussian distribution. We use this result for a probabilistic fit test, called the Gaussian Percentile Approximation (GPA), for standard bin-packing algorithms. To check whether a new task will fit into a machine, GPA checks whether the resulting distribution's percentile corresponding to the requested service level objective, SLO is still below the machine's capacity. In our simulation experiments, GPA resulted in colocations exceeding the machines' capacity with a frequency similar to the requested SLO.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79177233","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}
Mijung Kim, Jun Yu Li, Haris Volos, M. Marwah, A. Ulanov, K. Keeton, Joseph A. Tucek, L. Cherkasova, Le Xu, Pradeep R. Fernando
Given the growing availability of affordable scale-up servers, our goal is to bring the performance benefits of in-memory processing on scale-up servers to an increasingly common class of data analytics applications that process small to medium size datasets (up to a few 100GBs) that can easily fit in the memory of a typical scale-up server To achieve this goal, we leverage Spark, an existing memory-centric data analytics framework with wide-spread adoption among data scientists. Bringing Spark's data analytic capabilities to a scale-up system requires rethinking the original design assumptions, which, although effective for a scale-out system, are a poor match to a scale-up system resulting in unnecessary communication and memory inefficiencies.
{"title":"Sparkle: optimizing spark for large memory machines and analytics","authors":"Mijung Kim, Jun Yu Li, Haris Volos, M. Marwah, A. Ulanov, K. Keeton, Joseph A. Tucek, L. Cherkasova, Le Xu, Pradeep R. Fernando","doi":"10.1145/3127479.3134762","DOIUrl":"https://doi.org/10.1145/3127479.3134762","url":null,"abstract":"Given the growing availability of affordable scale-up servers, our goal is to bring the performance benefits of in-memory processing on scale-up servers to an increasingly common class of data analytics applications that process small to medium size datasets (up to a few 100GBs) that can easily fit in the memory of a typical scale-up server To achieve this goal, we leverage Spark, an existing memory-centric data analytics framework with wide-spread adoption among data scientists. Bringing Spark's data analytic capabilities to a scale-up system requires rethinking the original design assumptions, which, although effective for a scale-out system, are a poor match to a scale-up system resulting in unnecessary communication and memory inefficiencies.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"38 2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85770616","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}
In spite of the recent advancement of machine learning research, modern machine learning systems are still far from easy to use, at least from the perspective of business users or even scientists without a computer science background. Recently, there is a trend toward pushing machine learning onto the cloud as a "service," a.k.a. machine learning clouds. By putting a set of machine learning primitives on the cloud, these services significantly raise the level of abstraction for machine learning. For example, with Amazon Machine Learning, users only need to upload the dataset and specify the type of task (classification or regression). The cloud will then train machine learning models without any user intervention.
{"title":"How good are machine learning clouds for binary classification with good features?: extended abstract","authors":"Hantian Zhang, L. Zeng, Wentao Wu, Ce Zhang","doi":"10.1145/3127479.3132570","DOIUrl":"https://doi.org/10.1145/3127479.3132570","url":null,"abstract":"In spite of the recent advancement of machine learning research, modern machine learning systems are still far from easy to use, at least from the perspective of business users or even scientists without a computer science background. Recently, there is a trend toward pushing machine learning onto the cloud as a \"service,\" a.k.a. machine learning clouds. By putting a set of machine learning primitives on the cloud, these services significantly raise the level of abstraction for machine learning. For example, with Amazon Machine Learning, users only need to upload the dataset and specify the type of task (classification or regression). The cloud will then train machine learning models without any user intervention.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86629225","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 growing pressure on cloud application scalability has accentuated storage performance as a critical bottleneck. Although cache replacement algorithms have been extensively studied, cache prefetching - reducing latency by retrieving items before they are actually requested - remains an underexplored area. Existing approaches to history-based prefetching, in particular, provide too few benefits for real systems for the resources they cost. We propose Mithril, a prefetching layer that efficiently exploits historical patterns in cache request associations. Mithril is inspired by sporadic association rule mining and only relies on the timestamps of requests. Through evaluation of 135 block-storage traces, we show that Mithril is effective, giving an average of a 55% hit ratio increase over LRU and Probability Graph, and a 36% hit ratio gain over Amp at reasonable cost. Finally, we demonstrate the improvement comes from Mithril being able to capture mid-frequency blocks.
{"title":"Mithril: mining sporadic associations for cache prefetching","authors":"Juncheng Yang, Reza Karimi, Trausti Sæmundsson, Avani Wildani, Ymir Vigfusson","doi":"10.1145/3127479.3131210","DOIUrl":"https://doi.org/10.1145/3127479.3131210","url":null,"abstract":"The growing pressure on cloud application scalability has accentuated storage performance as a critical bottleneck. Although cache replacement algorithms have been extensively studied, cache prefetching - reducing latency by retrieving items before they are actually requested - remains an underexplored area. Existing approaches to history-based prefetching, in particular, provide too few benefits for real systems for the resources they cost. We propose Mithril, a prefetching layer that efficiently exploits historical patterns in cache request associations. Mithril is inspired by sporadic association rule mining and only relies on the timestamps of requests. Through evaluation of 135 block-storage traces, we show that Mithril is effective, giving an average of a 55% hit ratio increase over LRU and Probability Graph, and a 36% hit ratio gain over Amp at reasonable cost. Finally, we demonstrate the improvement comes from Mithril being able to capture mid-frequency blocks.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"28 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80212062","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}
Eric Jonas, Qifan Pu, S. Venkataraman, I. Stoica, B. Recht
Distributed computing remains inaccessible to a large number of users, in spite of many open source platforms and extensive commercial offerings. While distributed computation frameworks have moved beyond a simple map-reduce model, many users are still left to struggle with complex cluster management and configuration tools, even for running simple embarrassingly parallel jobs. We argue that stateless functions represent a viable platform for these users, eliminating cluster management overhead, fulfilling the promise of elasticity. Furthermore, using our prototype implementation, PyWren, we show that this model is general enough to implement a number of distributed computing models, such as BSP, efficiently. Extrapolating from recent trends in network bandwidth and the advent of disaggregated storage, we suggest that stateless functions are a natural fit for data processing in future computing environments.
{"title":"Occupy the cloud: distributed computing for the 99%","authors":"Eric Jonas, Qifan Pu, S. Venkataraman, I. Stoica, B. Recht","doi":"10.1145/3127479.3128601","DOIUrl":"https://doi.org/10.1145/3127479.3128601","url":null,"abstract":"Distributed computing remains inaccessible to a large number of users, in spite of many open source platforms and extensive commercial offerings. While distributed computation frameworks have moved beyond a simple map-reduce model, many users are still left to struggle with complex cluster management and configuration tools, even for running simple embarrassingly parallel jobs. We argue that stateless functions represent a viable platform for these users, eliminating cluster management overhead, fulfilling the promise of elasticity. Furthermore, using our prototype implementation, PyWren, we show that this model is general enough to implement a number of distributed computing models, such as BSP, efficiently. Extrapolating from recent trends in network bandwidth and the advent of disaggregated storage, we suggest that stateless functions are a natural fit for data processing in future computing environments.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":"24 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2017-02-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74510514","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：