Large volumes of raw data are naturally generated in multiple geographical regions. Modern data parallel frameworks may suffer from a degraded performance, due to the much lower inter-datacenter WAN bandwidth. We present Siphon, a new high-performance substrate that improves the ability of modern data parallel frameworks to optimize WAN transfers. Siphon aggregates inter-datacenter flows and routes them with the awareness of runtime network capacities. Following the principles of software-defined networking, Siphon is designed to accommodate a variety of flow scheduling algorithms that can improve the job level performance.
{"title":"Siphon: a high-performance substrate for inter-datacenter transfers in wide-area data analytics","authors":"Shuhao Liu, Li Chen, Baochun Li","doi":"10.1145/3127479.3132561","DOIUrl":"https://doi.org/10.1145/3127479.3132561","url":null,"abstract":"Large volumes of raw data are naturally generated in multiple geographical regions. Modern data parallel frameworks may suffer from a degraded performance, due to the much lower inter-datacenter WAN bandwidth. We present Siphon, a new high-performance substrate that improves the ability of modern data parallel frameworks to optimize WAN transfers. Siphon aggregates inter-datacenter flows and routes them with the awareness of runtime network capacities. Following the principles of software-defined networking, Siphon is designed to accommodate a variety of flow scheduling algorithms that can improve the job level performance.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89556285","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}
Dan Huang, J. Wang, Qing Liu, Xuhong Zhang, Xunchao Chen, Jian Zhou
Today BigData systems commonly use resource management systems such as TORQUE, Mesos, and Google Borg to share the physical resources among users or applications. Enabled by virtualization, users can run their applications on the same node with low mutual interference. Container-based virtualizations (e.g., Docker and Linux Containers) offer a lightweight virtualization layer, which promises a near-native performance and is adopted by some Big-Data resource sharing platforms such as Mesos. Nevertheless, using containers to consolidate the I/O resources of shared storage systems is still at an early stage, especially in a distributed file system (DFS) such as Hadoop File System (HDFS). To overcome this issue, we propose a distributed middleware system, DFS-Container, by further containerizing DFS. We also evaluate and analyze the unfairness of using containers to proportionally allocate the I/O resource of DFS. Based on these analyses and evaluations, we propose and implement a new mechanism, IOPS-Regulator, which improve the fairness of proportional allocation by 74.4% on average.
{"title":"DFS-container: achieving containerized block I/O for distributed file systems","authors":"Dan Huang, J. Wang, Qing Liu, Xuhong Zhang, Xunchao Chen, Jian Zhou","doi":"10.1145/3127479.3132568","DOIUrl":"https://doi.org/10.1145/3127479.3132568","url":null,"abstract":"Today BigData systems commonly use resource management systems such as TORQUE, Mesos, and Google Borg to share the physical resources among users or applications. Enabled by virtualization, users can run their applications on the same node with low mutual interference. Container-based virtualizations (e.g., Docker and Linux Containers) offer a lightweight virtualization layer, which promises a near-native performance and is adopted by some Big-Data resource sharing platforms such as Mesos. Nevertheless, using containers to consolidate the I/O resources of shared storage systems is still at an early stage, especially in a distributed file system (DFS) such as Hadoop File System (HDFS). To overcome this issue, we propose a distributed middleware system, DFS-Container, by further containerizing DFS. We also evaluate and analyze the unfairness of using containers to proportionally allocate the I/O resource of DFS. Based on these analyses and evaluations, we propose and implement a new mechanism, IOPS-Regulator, which improve the fairness of proportional allocation by 74.4% on average.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87024643","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}
Online services are often deployed over geographically-scattered data centers (geo-replication), which allows services to be highly available and reduces access latency. On the down side, to provide ACID transactions, global certification (i.e., across data centers) is needed to detect conflicts between concurrent transactions executing at different data centers. The global certification phase reduces throughput because transactions need to hold pre-commit locks, and it increases client-perceived latency because global certification lies in the critical path of transaction execution.
{"title":"Exploiting speculation in partially replicated transactional data stores","authors":"Zhongmiao Li, P. V. Roy, P. Romano","doi":"10.1145/3127479.3132692","DOIUrl":"https://doi.org/10.1145/3127479.3132692","url":null,"abstract":"Online services are often deployed over geographically-scattered data centers (geo-replication), which allows services to be highly available and reduces access latency. On the down side, to provide ACID transactions, global certification (i.e., across data centers) is needed to detect conflicts between concurrent transactions executing at different data centers. The global certification phase reduces throughput because transactions need to hold pre-commit locks, and it increases client-perceived latency because global certification lies in the critical path of transaction execution.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88082394","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}
M. Karpathiotakis, A. Floratou, Fatma Özcan, A. Ailamaki
The typical enterprise data architecture consists of several actively updated data sources (e.g., NoSQL systems, data warehouses), and a central data lake such as HDFS, in which all the data is periodically loaded through ETL processes. To simplify query processing, state-of-the-art data analysis approaches solely operate on top of the local, historical data in the data lake, and ignore the fresh tail end of data that resides in the original remote sources. However, as many business operations depend on real-time analytics, this approach is no longer viable. The alternative is hand-crafting the analysis task to explicitly consider the characteristics of the various data sources and identify optimization opportunities, rendering the overall analysis non-declarative and convoluted. Based on our experiences operating in data lake environments, we design System-PV, a real-time analytics system that masks the complexity of dealing with multiple data sources while offering minimal response times. System-PV extends Spark with a sophisticated data virtualization module that supports multiple applications - from SQL queries to machine learning. The module features a location-aware compiler that considers source complexity, and a two-phase optimizer that produces and refines the query plans, not only for SQL queries but for all other types of analysis as well. The experiments show that System-PV is often faster than Spark by more than an order of magnitude. In addition, the experiments show that the approach of accessing both the historical and the remote fresh data is viable, as it performs comparably to solely operating on top of the local, historical data.
{"title":"No data left behind: real-time insights from a complex data ecosystem","authors":"M. Karpathiotakis, A. Floratou, Fatma Özcan, A. Ailamaki","doi":"10.1145/3127479.3131208","DOIUrl":"https://doi.org/10.1145/3127479.3131208","url":null,"abstract":"The typical enterprise data architecture consists of several actively updated data sources (e.g., NoSQL systems, data warehouses), and a central data lake such as HDFS, in which all the data is periodically loaded through ETL processes. To simplify query processing, state-of-the-art data analysis approaches solely operate on top of the local, historical data in the data lake, and ignore the fresh tail end of data that resides in the original remote sources. However, as many business operations depend on real-time analytics, this approach is no longer viable. The alternative is hand-crafting the analysis task to explicitly consider the characteristics of the various data sources and identify optimization opportunities, rendering the overall analysis non-declarative and convoluted. Based on our experiences operating in data lake environments, we design System-PV, a real-time analytics system that masks the complexity of dealing with multiple data sources while offering minimal response times. System-PV extends Spark with a sophisticated data virtualization module that supports multiple applications - from SQL queries to machine learning. The module features a location-aware compiler that considers source complexity, and a two-phase optimizer that produces and refines the query plans, not only for SQL queries but for all other types of analysis as well. The experiments show that System-PV is often faster than Spark by more than an order of magnitude. In addition, the experiments show that the approach of accessing both the historical and the remote fresh data is viable, as it performs comparably to solely operating on top of the local, historical data.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86290114","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}
Next-generation non-volatile memories (NVMs) will provide byte addressability, persistence, high density, and DRAM-like performance. They have the potential to benefit many datacenter applications. However, most previous research on NVMs has focused on using them in a single machine environment. It is still unclear how to best utilize them in distributed, datacenter environments. We introduce Distributed Shared Persistent Memory (DSPM), a new framework for using persistent memories in distributed data-center environments. DSPM provides a new abstraction that allows applications to both perform traditional memory load and store instructions and to name, share, and persist their data. We built Hotpot, a kernel-level DSPM system that provides low-latency, transparent memory accesses, data persistence, data reliability, and high availability. The key ideas of Hotpot are to integrate distributed memory caching and data replication techniques and to exploit application hints. We implemented Hotpot in the Linux kernel and demonstrated its benefits by building a distributed graph engine on Hotpot and porting a NoSQL database to Hotpot. Our evaluation shows that Hotpot outperforms a recent distributed shared memory system by 1.3× to 3.2× and a recent distributed PM-based file system by 1.5× to 3.0×.
{"title":"Distributed shared persistent memory","authors":"Yizhou Shan, Shin-Yeh Tsai, Yiying Zhang","doi":"10.1145/3127479.3128610","DOIUrl":"https://doi.org/10.1145/3127479.3128610","url":null,"abstract":"Next-generation non-volatile memories (NVMs) will provide byte addressability, persistence, high density, and DRAM-like performance. They have the potential to benefit many datacenter applications. However, most previous research on NVMs has focused on using them in a single machine environment. It is still unclear how to best utilize them in distributed, datacenter environments. We introduce Distributed Shared Persistent Memory (DSPM), a new framework for using persistent memories in distributed data-center environments. DSPM provides a new abstraction that allows applications to both perform traditional memory load and store instructions and to name, share, and persist their data. We built Hotpot, a kernel-level DSPM system that provides low-latency, transparent memory accesses, data persistence, data reliability, and high availability. The key ideas of Hotpot are to integrate distributed memory caching and data replication techniques and to exploit application hints. We implemented Hotpot in the Linux kernel and demonstrated its benefits by building a distributed graph engine on Hotpot and porting a NoSQL database to Hotpot. Our evaluation shows that Hotpot outperforms a recent distributed shared memory system by 1.3× to 3.2× and a recent distributed PM-based file system by 1.5× to 3.0×.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73602956","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}
Qizhen Zhang, Hongzhi Chen, D. Yan, James Cheng, B. T. Loo, P. Bangalore
Graph analytics systems have gained significant popularity due to the prevalence of graph data. Many of these systems are designed to run in a shared-nothing architecture whereby a cluster of machines can process a large graph in parallel. In more recent proposals, others have argued that a single-machine system can achieve better performance and/or is more cost-effective. There is however no clear consensus which approach is better. In this paper, we classify existing graph analytics systems into four categories based on the architectural differences, i.e., processing infrastructure (centralized vs distributed), and memory consumption (in-memory vs out-of-core). We select eight open-source systems to cover all categories, and perform a comparative measurement study to compare their performance and cost characteristics across a spectrum of input data, applications, and hardware settings. Our results show that the best performing configuration can depend on the type of applications and input graphs, and there is no dominant winner across all categories. Based on our findings, we summarize the trends in performance and cost, and provide several insights that help to illuminate the performance and resource cost tradeoffs across different graph analytics systems and categories.
{"title":"Architectural implications on the performance and cost of graph analytics systems","authors":"Qizhen Zhang, Hongzhi Chen, D. Yan, James Cheng, B. T. Loo, P. Bangalore","doi":"10.1145/3127479.3128606","DOIUrl":"https://doi.org/10.1145/3127479.3128606","url":null,"abstract":"Graph analytics systems have gained significant popularity due to the prevalence of graph data. Many of these systems are designed to run in a shared-nothing architecture whereby a cluster of machines can process a large graph in parallel. In more recent proposals, others have argued that a single-machine system can achieve better performance and/or is more cost-effective. There is however no clear consensus which approach is better. In this paper, we classify existing graph analytics systems into four categories based on the architectural differences, i.e., processing infrastructure (centralized vs distributed), and memory consumption (in-memory vs out-of-core). We select eight open-source systems to cover all categories, and perform a comparative measurement study to compare their performance and cost characteristics across a spectrum of input data, applications, and hardware settings. Our results show that the best performing configuration can depend on the type of applications and input graphs, and there is no dominant winner across all categories. Based on our findings, we summarize the trends in performance and cost, and provide several insights that help to illuminate the performance and resource cost tradeoffs across different graph analytics systems and categories.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90180627","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}
Modern in-memory distributed computation frameworks like Spark adequately leverage memory resources to cache intermediate data across multi-stage tasks in pre-allocated worker processes, so as to speedup executions. They rely on a cluster resource manager like Yarn or Mesos to pre-reserve specific amount of CPU and memory for workers ahead of task scheduling. Since a worker is executed for an entire application and runs multiple batches of DAG tasks from multi-stages, its memory demands change over time [3]. Resource managers like Yarn solve the non-trivial allocation problem of determining right amounts of memory provision for workers by requiring users to make explicit reservations before execution. Since the underlying execution frameworks, workload and complex codebases are invisible, users tend to over-estimate or under-estimate workers' demands, leading to over-provisioning or under-provisioning of memory resources. We observed there exists a performance inflection point with respect to memory reservation per stage of applications. After that, performance fluctuates little even under over-provisioned memory [1]. It is the minimum required memory to achieve expected nearly optimal performance. We call these capacities as optimal demands. They are capacity cut lines to divide over-provisioning and under-provisioning. To relieve the burden of users, and provide guarantees over both maximum cluster memory utilization and optimal application performance, we present a system namely Prometheus for online estimation of optimal memory demand for workers per future stage, without involving users' efforts. The procedure to explore optimal demands is essentially a search problem correlated memory reservation and performance. Most existing searching methods [2] need multiple profiling runs or prior historical execution statistics, which are not applicable to online configuration of newly submitted or non-recurring jobs. The recurring applications' optimal demands also change over time under variations of input datasets, algorithmic parameters or source code. It becomes too expensive and infeasible to rebuild new search model for every setting. Prometheus adopts a two-step approach to tackle the problem: 1) For newly submitted or non-recurring jobs, we do profiling and histogram frequency analysis of job's runtime memory footprints from only one pilot run under over-provisioned memory. It achieves a highly accurate (over 80% accuracy) initial estimation of optimal demands per stage for each worker. By analyzing frequency of past memory usages per sampling time, we efficiently estimate probability of base demands and distinguish them from unnecessarily excessive usages. Allocation of base demands tends to achieve near-optimal performance, so as to approach optimal demands. 2) Histogram frequency analysis algorithm has an intrinsic property of self-decay. For subsequent recurring submissions, Prometheus exploits this property to efficiently perform a recur
{"title":"Prometheus: online estimation of optimal memory demands for workers in in-memory distributed computation","authors":"Guoyao Xu, Chengzhong Xu","doi":"10.1145/3127479.3132689","DOIUrl":"https://doi.org/10.1145/3127479.3132689","url":null,"abstract":"Modern in-memory distributed computation frameworks like Spark adequately leverage memory resources to cache intermediate data across multi-stage tasks in pre-allocated worker processes, so as to speedup executions. They rely on a cluster resource manager like Yarn or Mesos to pre-reserve specific amount of CPU and memory for workers ahead of task scheduling. Since a worker is executed for an entire application and runs multiple batches of DAG tasks from multi-stages, its memory demands change over time [3]. Resource managers like Yarn solve the non-trivial allocation problem of determining right amounts of memory provision for workers by requiring users to make explicit reservations before execution. Since the underlying execution frameworks, workload and complex codebases are invisible, users tend to over-estimate or under-estimate workers' demands, leading to over-provisioning or under-provisioning of memory resources. We observed there exists a performance inflection point with respect to memory reservation per stage of applications. After that, performance fluctuates little even under over-provisioned memory [1]. It is the minimum required memory to achieve expected nearly optimal performance. We call these capacities as optimal demands. They are capacity cut lines to divide over-provisioning and under-provisioning. To relieve the burden of users, and provide guarantees over both maximum cluster memory utilization and optimal application performance, we present a system namely Prometheus for online estimation of optimal memory demand for workers per future stage, without involving users' efforts. The procedure to explore optimal demands is essentially a search problem correlated memory reservation and performance. Most existing searching methods [2] need multiple profiling runs or prior historical execution statistics, which are not applicable to online configuration of newly submitted or non-recurring jobs. The recurring applications' optimal demands also change over time under variations of input datasets, algorithmic parameters or source code. It becomes too expensive and infeasible to rebuild new search model for every setting. Prometheus adopts a two-step approach to tackle the problem: 1) For newly submitted or non-recurring jobs, we do profiling and histogram frequency analysis of job's runtime memory footprints from only one pilot run under over-provisioned memory. It achieves a highly accurate (over 80% accuracy) initial estimation of optimal demands per stage for each worker. By analyzing frequency of past memory usages per sampling time, we efficiently estimate probability of base demands and distinguish them from unnecessarily excessive usages. Allocation of base demands tends to achieve near-optimal performance, so as to approach optimal demands. 2) Histogram frequency analysis algorithm has an intrinsic property of self-decay. For subsequent recurring submissions, Prometheus exploits this property to efficiently perform a recur","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77096187","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}
Cheng Wang, Jianyu Jiang, Xusheng Chen, Ning Yi, Heming Cui
State machine replication (SMR) uses Paxos to enforce the same inputs for a program (e.g., Redis) replicated on a number of hosts, tolerating various types of failures. Unfortunately, traditional Paxos protocols incur prohibitive performance overhead on server programs due to their high consensus latency on TCP/IP. Worse, the consensus latency of extant Paxos protocols increases drastically when more concurrent client connections or hosts are added. This paper presents APUS, the first RDMA-based Paxos protocol that aims to be fast and scalable to client connections and hosts. APUS intercepts inbound socket calls of an unmodified server program, assigns a total order for all input requests, and uses fast RDMA primitives to replicate these requests concurrently. We evaluated APUS on nine widely-used server programs (e.g., Redis and MySQL). APUS incurred a mean overhead of 4.3% in response time and 4.2% in throughput. We integrated APUS with an SMR system Calvin. Our Calvin-APUS integration was 8.2X faster than the extant Calvin-ZooKeeper integration. The consensus latency of APUS outperformed an RDMA-based consensus protocol by 4.9X. APUS source code and raw results are released on github.com/hku-systems/apus.
{"title":"APUS: fast and scalable paxos on RDMA","authors":"Cheng Wang, Jianyu Jiang, Xusheng Chen, Ning Yi, Heming Cui","doi":"10.1145/3127479.3128609","DOIUrl":"https://doi.org/10.1145/3127479.3128609","url":null,"abstract":"State machine replication (SMR) uses Paxos to enforce the same inputs for a program (e.g., Redis) replicated on a number of hosts, tolerating various types of failures. Unfortunately, traditional Paxos protocols incur prohibitive performance overhead on server programs due to their high consensus latency on TCP/IP. Worse, the consensus latency of extant Paxos protocols increases drastically when more concurrent client connections or hosts are added. This paper presents APUS, the first RDMA-based Paxos protocol that aims to be fast and scalable to client connections and hosts. APUS intercepts inbound socket calls of an unmodified server program, assigns a total order for all input requests, and uses fast RDMA primitives to replicate these requests concurrently. We evaluated APUS on nine widely-used server programs (e.g., Redis and MySQL). APUS incurred a mean overhead of 4.3% in response time and 4.2% in throughput. We integrated APUS with an SMR system Calvin. Our Calvin-APUS integration was 8.2X faster than the extant Calvin-ZooKeeper integration. The consensus latency of APUS outperformed an RDMA-based consensus protocol by 4.9X. APUS source code and raw results are released on github.com/hku-systems/apus.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89181489","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 Radio Access Networks (CRAN) has been proposed for 5G networks for better flexibility, scalability, and performance. CRAN aims to apply cloud-like architecture on RAN. This paper focuses on cache management and load balance in the software architecture of CRAN, intends to provide fault mitigation and carrier-grade real-time services. First, a new cache management algorithm based on event frequency and QoS level of User Equipment (UE) is proposed, utilizing Exponential Decay scoring function and Analytic Hierarchy Process. Next, load balance (LB) function is added, and five LB algorithms are evaluated. The performance evaluation is based on real-life UE event characteristics and RAN system values provided by Nokia Research. Results show the cache hit rate, delay and network traffic; as well as the queue-length analysis of LB algorithms.
{"title":"Abstract: cache management and load balancing for 5G cloud radio access networks","authors":"Chin Tsai, M. Moh","doi":"10.1145/3127479.3132690","DOIUrl":"https://doi.org/10.1145/3127479.3132690","url":null,"abstract":"Cloud Radio Access Networks (CRAN) has been proposed for 5G networks for better flexibility, scalability, and performance. CRAN aims to apply cloud-like architecture on RAN. This paper focuses on cache management and load balance in the software architecture of CRAN, intends to provide fault mitigation and carrier-grade real-time services. First, a new cache management algorithm based on event frequency and QoS level of User Equipment (UE) is proposed, utilizing Exponential Decay scoring function and Analytic Hierarchy Process. Next, load balance (LB) function is added, and five LB algorithms are evaluated. The performance evaluation is based on real-life UE event characteristics and RAN system values provided by Nokia Research. Results show the cache hit rate, delay and network traffic; as well as the queue-length analysis of LB algorithms.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89898515","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}
Many cloud-based data management and analytics systems support complex objects. Dataflow platforms such as Spark and Flink allow programmers to manipulate sets consisting of objects from a host programming language (often Java). Document databases such as MongoDB make use of hierarchical interchange formats---most popularly JSON---which embody a data model where individual records can themselves contain sets of records. Systems such as Dremel and AsterixDB allow complex nesting of data structures. Clearly, no system designer would expect a system that stores JSON objects as text to perform at the same level as a system based upon a custom-built physical data model. The question we ask is: How significant is the performance hit associated with choosing a particular physical implementation? Is the choice going to result in a negligible performance cost, or one that is debilitating? Unfortunately, there does not exist a scientific study of the effect of physical complex model implementation on system performance in the literature. Hence it is difficult for a system designer to fully understand performance implications of such choices. This paper is an attempt to remedy that.
{"title":"An experimental comparison of complex object implementations for big data systems","authors":"Sourav Sikdar, Kia Teymourian, C. Jermaine","doi":"10.1145/3127479.3129248","DOIUrl":"https://doi.org/10.1145/3127479.3129248","url":null,"abstract":"Many cloud-based data management and analytics systems support complex objects. Dataflow platforms such as Spark and Flink allow programmers to manipulate sets consisting of objects from a host programming language (often Java). Document databases such as MongoDB make use of hierarchical interchange formats---most popularly JSON---which embody a data model where individual records can themselves contain sets of records. Systems such as Dremel and AsterixDB allow complex nesting of data structures. Clearly, no system designer would expect a system that stores JSON objects as text to perform at the same level as a system based upon a custom-built physical data model. The question we ask is: How significant is the performance hit associated with choosing a particular physical implementation? Is the choice going to result in a negligible performance cost, or one that is debilitating? Unfortunately, there does not exist a scientific study of the effect of physical complex model implementation on system performance in the literature. Hence it is difficult for a system designer to fully understand performance implications of such choices. This paper is an attempt to remedy that.","PeriodicalId":20679,"journal":{"name":"Proceedings of the 2017 Symposium on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2017-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84148719","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：