{"title":"Running Interactive Perception Applications on Open Cirrus","authors":"Qian Zhu, N. Yigitbasi, P. Pillai","doi":"10.1109/OCS.2011.9","DOIUrl":"https://doi.org/10.1109/OCS.2011.9","url":null,"abstract":"Interactive perception applications, such as gesture recognition and vision-based user interfaces, process high-data rate streams with compute intensive computer vision and machine learning algorithms. Yet, they require extremely low latencies to remain interactive and ensure timely results to users. Cluster computing resources, such as those provided by Open Cirrus deployments, can help address the computation requirements, but significant challenges exist in practice. This paper highlights our efforts to parallelize interactive perception applications, tune them for best fidelity and latency, and place, schedule, and execute them on a cluster platform. We also look at remaining open problems and potential solutions.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127849221","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}
Zhenghua Xue, Jianhui Li, Yuanchun Zhou, Yang Zhang, Geng Shen
{"title":"An Open Job Scheduling Service for Large-Scale Data Processing","authors":"Zhenghua Xue, Jianhui Li, Yuanchun Zhou, Yang Zhang, Geng Shen","doi":"10.1109/OCS.2011.11","DOIUrl":"https://doi.org/10.1109/OCS.2011.11","url":null,"abstract":"Scientific Data Grid (SDG) of Chinese Academy of Sciences aims at integrating distributed scientific data, providing transparent data access mechanism and efficient data analysis, processing and visualization services. SDG Job Scheduler (SDGJS) adopts an open and service-oriented framework. The scheduling policy of SDGJS considers both performance of computing nodes and distribution of large data so that it can achieve an efficient job processing. SDGJS has successfully served for some data grid applications.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"8 5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129636899","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":"Evaluation of HPC Applications on Cloud","authors":"Abhishek K. Gupta, D. Milojicic","doi":"10.1109/OCS.2011.10","DOIUrl":"https://doi.org/10.1109/OCS.2011.10","url":null,"abstract":"HPC applications are increasingly being used in academia and laboratories for scientific research and in industries for business and analytics. Cloud computing offers the benefits of virtualization, elasticity of resources and elimination of cluster setup cost and time to HPC applications users. However, poor network performance, performance variation and OS noise are some of the challenges for execution of HPC applications on Cloud. In this paper, we propose that Cloud can be viable platform for some HPC applications depending upon application characteristics such as communication volume and pattern and sensitivity to OS noise and scale. We present an evaluation of the performance and cost tradeoffs of HPC applications on a range of platforms varying from Cloud (with and without virtualization) to HPC-optimized cluster. Our results show that Cloud is viable platform for some applications, specifically, non communicationintensive applications such as embarrassingly parallel and tree-structured computations up to high processor count and for communication-intensive applications up to low processor count.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"13 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122361267","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":"The Application and Practice of Parallel Cloud Computing in ISP","authors":"Zhi-Xian Huang, Guoliang Yang, Shengyong Ding","doi":"10.1109/OCS.2011.7","DOIUrl":"https://doi.org/10.1109/OCS.2011.7","url":null,"abstract":"In order to deal with massive data processing, China Telecom has deployed a platform named distributed service engine, which is based on parallel cloud computing technology. On this platform, we also developed some tele-data processing system and internet applications to show the effectiveness of this platform. This paper describes the design of this platform and how it can be used in telecom industry.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125116211","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}
Anshul Gandhi, Mor Harchol-Balter, R. Raghunathan, M. Kozuch
{"title":"Distributed, Robust Auto-Scaling Policies for Power Management in Compute Intensive Server Farms","authors":"Anshul Gandhi, Mor Harchol-Balter, R. Raghunathan, M. Kozuch","doi":"10.1109/OCS.2011.6","DOIUrl":"https://doi.org/10.1109/OCS.2011.6","url":null,"abstract":"Server farms today often over-provision resources to handle peak demand, resulting in an excessive waste of power. Ideally, server farm capacity should be dynamically adjusted based on the incoming demand. However, the unpredictable and time-varying nature of customer demands makes it very difficult to efficiently scale capacity in server farms. The problem is further exacerbated by the large setup time needed to increase capacity, which can adversely impact response times as well as utilize additional power.In this paper, we present the design and implementation of a class of Distributed and Robust Auto-Scaling policies (DRAS policies), for power management in compute intensive server farms. Results indicate that the DRAS policies dynamically adjust server farm capacity without requiring any prediction of the future load, or any feedback control. Implementation results on a 21 server test-bed show that the DRAS policies provide near-optimal response time while lowering power consumption by about 30% when compared to static provisioning policies that employ a fixed number of servers.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125698932","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}
Sudarsun Kannan, D. Milojicic, V. Talwar, Ada Gavrilovska, K. Schwan, H. Abbasi
{"title":"Using Active NVRAM for Cloud I/O","authors":"Sudarsun Kannan, D. Milojicic, V. Talwar, Ada Gavrilovska, K. Schwan, H. Abbasi","doi":"10.1109/OCS.2011.12","DOIUrl":"https://doi.org/10.1109/OCS.2011.12","url":null,"abstract":"A well-known problem for large scale cloud applications is how to scale their I/O performance. While next generation storage class memories like phase change memory and Memristors offer potential for high I/O bandwidths, if left unchecked, the raw volumes and rates of I/O already present in current cloud applications can quickly overwhelm future I/O infrastructures. This fact is motivating research on 'data staging' in which I/O and data movement actions are enhanced with computations that process data before or while moving it across I/O channels -- in situ -- to filter or reduce it, to better organize it for subsequent access (e.g., by other applications as in coupled codes), or to analyze it to quickly derive important insights about the application producing those large data volumes. This paper proposes a technique that uses and exploits 'Active NVRAM' (non volatile memory) for staging I/O. Active NVRAMs are node-local NVRAMs that are embedded with a low power system-on-chip compute element. These active compute elements can be used to operate on output data asynchronously with the tasks performed by computational node elements, to reduce data or to perform some of the data processing required for data analytics before data is moved to longer term storage. The paper describes the Active NVRAM design, sample ways in which it is used for I/O acceleration, and initial performance results evaluating the opportunities for and limitations of the Active NVRAM approach.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116015595","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}
Prakash Reddy, Shariff Dudekula, Susanth Puthanveedu, D. Milojicic
{"title":"Globally Distributed BookPrep - Open Crirrus-Hosted Service for Book Preparation","authors":"Prakash Reddy, Shariff Dudekula, Susanth Puthanveedu, D. Milojicic","doi":"10.1109/OCS.2011.8","DOIUrl":"https://doi.org/10.1109/OCS.2011.8","url":null,"abstract":"BookPrep is a Print-On-Demand service that takes raw scans and converts them to print-ready files. It requires large amount of storage and takes an average of 5 hours of CPU time to process a single book with about 300 pages. The experiment we conducted involved moving the processing of books on Open Cirrus closer to the location of the data. At three Open Cirrus sites we installed BookPrep service and we pre-populated each site with region-specific scanned books. When requests come in to process a book, each request is routed to the compute node closest to the source data. The compute node is then expected to store the processed data on the same network. The compute nodes are allocated and deallocated based on demand. There is a cloud based metadata repository that is used to update the metadata associated with each book regardless of the location of the source and derived data. The goal of this experiment is to determine if performance can be improved by moving book processing close closer to source data location. The fundamental reason behind the success of MapReduce is the notion of moving compute close to data and we would like to see if that same principal can be applied to a pull based scheduling model.","PeriodicalId":346897,"journal":{"name":"2011 Sixth Open Cirrus Summit","volume":"109 1-2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120996951","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}