Spark Job Research Articles

Energy-aware scheduling for spark job based on deep reinforcement learning in cloud

Energy-aware scheduling for spark job based on deep reinforcement learning in cloud

Performance and Cost-Efficient Spark Job Scheduling Based on Deep Reinforcement Learning in Cloud Computing Environments

Big data frameworks such as Spark and Hadoop are widely adopted to run analytics jobs in both research and industry. Cloud offers affordable compute resources which are easier to manage. Hence, many organizations are shifting towards a cloud deployment of their big data computing clusters. However, job scheduling is a complex problem in the presence of various Service Level Agreement (SLA) objectives such as monetary cost reduction, and job performance improvement. Most of the existing research does not address multiple objectives together and fail to capture the inherent cluster and workload characteristics. In this article, we formulate the job scheduling problem of a cloud-deployed Spark cluster and propose a novel Reinforcement Learning (RL) model to accommodate the SLA objectives. We develop the RL cluster environment and implement two Deep Reinforce Learning (DRL) based schedulers in TF-Agents framework. The proposed DRL-based scheduling agents work at a fine-grained level to place the executors of jobs while leveraging the pricing model of cloud VM instances. In addition, the DRL-based agents can also learn the inherent characteristics of different types of jobs to find a proper placement to reduce both the total cluster VM usage cost and the average job duration. The results show that the proposed DRL-based algorithms can reduce the VM usage cost up to 30%.

Research on the Allocation Method of Regional Science and Technology Resources from the Perspective of Rationality

At present, the allocation efficiency of regional scientific and technological resources is low, and there are few research studies on social equity and economic efficiency. Therefore, this paper puts forward the allocation method of regional scientific and technological resources based on rationality. With the support of rationality perspective, the evaluation model of reform path of regional scientific and technological resource allocation is constructed to analyze the economic benefits and equity benefits of regional scientific and technological resource allocation. According to the principle of optimum allocation of regional science and technology resources, three-dimensional structure is constructed to maximize national investment and benefit and determine the optimal Pareto of resource allocation to measure the efficiency of resource allocation. The evaluation index system of the reform path of resource allocation is constructed by selecting the evaluation index of the reform path of resource allocation. The benchmark platform of big data was selected to generate data sets to be processed, and the spark on yam platform was used to submit jobs and generate spark job running data sets. The operation performance prediction model was established to optimize the configuration parameters of regional science and technology resources. The analysis results show that the designed method has high configuration capability and good effectiveness.

Deadline-Aware Cost Optimization for Spark

We present OptEx, a closed-form model of job execution on Apache Spark, a popular parallel processing engine. To the best of our knowledge, OptEx is the first work that analytically models job completion time on Spark. The model can be used to estimate the completion time of a given Spark job on a cloud, with respect to the size of the input dataset, the number of iterations, and the number of nodes comprising the underlying cluster. Experimental results demonstrate that OptEx yields a mean relative error of 6 percent in estimating the job completion time. Furthermore, the model can be applied for estimating the cost-optimal cluster composition for running a given Spark job on a cloud under a completion deadline specified in the SLO (i.e., Service Level Objective). We show experimentally that OptEx is able to correctly estimate the required cluster composition for running a given Spark job under a given SLO deadline with an accuracy of 98 percent. We also provide a tool which can classify Spark jobs into job categories based on bisimilarity analysis on lineage graphs collected from the given jobs.

Outlier Detection at GBIF Using DBSCAN

Geographic outliers at GBIF (Global Biodiversity Information Facility) are a known problem. Outliers can be errors, coordinates with high uncertainty, or simply occurrences from an undersampled region. Often in data cleaning pipelines, outliers are removed (even if they are legitimate points) because the researcher does not have time to verify each record one-by-one. Outlier points are usually occurrences that need attention. Currently, there is no outlier detection implemented at GBIF and it is up to the user to flag outliers themselves. DBSCAN (a density-based algorithm for discovering clusters in large spatial databases with noise) is a simple and popular clustering algorithm. It uses two parameters, (1) distance and (2) a minimum number of points per cluster, to decide if something is an outlier. Since occurrence data can be very patchy, non-clustering distance-based methods will fail often Fig. 1. DBSCAN does not need to know the expected number of clusters in advance. DBSCAN does well using only distance and does not require some additional environmental variables like Bioclim. Advanatages of DBSCAN : Simple Easy to understand Only two parameters to set Scales well No additional data sources needed Users would understand how their data was changed Simple Easy to understand Only two parameters to set Scales well No additional data sources needed Users would understand how their data was changed Drawbacks : Only uses distance Must choose parameter settings Sensitive to sparse global sampling Does not include any other relevant environmental information Can only flag outliers outside of a point blob Only uses distance Must choose parameter settings Sensitive to sparse global sampling Does not include any other relevant environmental information Can only flag outliers outside of a point blob Outlier detection and error detection are different. If your goal is to produce a system with no false positives, it will fail. While more complex environmentally-informed outlier detection methods (like reverse jackknifing (Chapman 2005)) might perform better for certain examples or even in genreal, DBSCAN performs adequately on almost everything despite being very simple. Currently I am using DBSCAN to find errors and assess dataset quality. It is a Spark job written in Scala (github). It does not run on species with lots of (>30K) unique latitude-longitude points, since the current implementation relies on an in-memory distance matrix. However, around 99% of species (plants, animals, fungi) on GBIF have fewer than >30K unique lat-long points (2,283 species keys / 222,993 species keys). There are other implementations ( example) that might scale to many more points. There are no immediate plans to include DBSCAN outliers as a data quality flag on GBIF, but it could be done somewhat easily, since this type of method does not rely on any external environmental data sources and already runs on the GBIF cluster.

Balance resource allocation for spark jobs based on prediction of the optimal resource

Apache Spark provides a well-known MapReduce computing framework, aiming to fast-process big data analytics in data-parallel manners. With this platform, large input data are divided into data partitions. Each data partition is processed by multiple computation tasks concurrently. Outputs of these computation tasks are transferred among multiple computers via the network. However, such a distributed computing framework suffers from system overheads, inevitably caused by communication and disk I/O operations. System overheads take up a large proportion of the Job Completion Time (JCT). We observed that excessive computational resources incurs considerable system overheads, prolonging the JCT. The over-allocation of individual jobs not only prolongs their own JCTs, but also likely makes other jobs suffer from under-allocation. Thus, the average JCT is suboptimal, too. To address this problem, we propose a prediction model to estimate the changing JCT of a single Spark job. With the support of the prediction method, we designed a heuristic algorithm to balance the resource allocation of multiple Spark jobs, aiming to minimize the average JCT in multiple-job cases. We implemented the prediction model and resource allocation method in ReB, a Resource-Balancer based on Apache Spark. Experimental results showed that ReB significantly outperformed the traditional max-min fairness and shortest-job-optimal methods. The average JCT was decreased by around 10%-30% compared to the existing solutions.

A gray-box modeling methodology for runtime prediction of Apache Spark jobs

Apache Spark jobs are often characterized by processing huge data sets and, therefore, require runtimes in the range of minutes to hours. Thus, being able to predict the runtime of such jobs would be useful not only to know when the job will finish, but also for scheduling purposes, to estimate monetary costs for cloud deployment, or to determine an appropriate cluster configuration, such as the number of nodes. However, predicting Spark job runtimes is much more challenging than for standard database queries: cluster configuration and parameters have a significant performance impact and jobs usually contain a lot of user-defined code making it difficult to estimate cardinalities and execution costs. In this paper, we present a gray-box modeling methodology for runtime prediction of Apache Spark jobs. Our approach comprises two steps: first, a white-box model for predicting the cardinalities of the input RDDs of each operator is built based on prior knowledge about the behavior and application parameters such as applied filters data, number of iterations, etc. In the second step, a black-box model for each task constructed by monitoring runtime metrics while varying allocated resources and input RDD cardinalities is used. We further show how to use this gray-box approach not only for predicting the runtime of a given job, but also as part of a decision model for reusing intermediate cached results of Spark jobs. Our methodology is validated with experimental evaluation showing a highly accurate prediction of the actual job runtime and a performance improvement if intermediate results can be reused.

Erratic server behavior detection using machine learning on streams of monitoring data

With the explosion of the number of distributed applications, a new dynamic server environment emerged grouping servers into clusters, utilization of which depends on the current demand for the application. To provide reliable and smooth services it is crucial to detect and fix possible erratic behavior of individual servers in these clusters. Use of standard techniques for this purpose requires manual work and delivers sub-optimal results. Using only application agnostic monitoring metrics our machine learning based method analyzes the recent performance of the inspected server as well as the state of the rest of the cluster, thus checking not only the behavior of the single server, but the load on the whole distributed application as well. We have implemented our method in a Spark job running in the CERN MONIT infrastructure. In this contribution we present results of testing multiple machine learning algorithms and pre-processing techniques to identify the servers erratic behavior. We also discuss the challenges of deploying our new method into production.

A rewrite-based optimizer for Spark

Spark is the leading platform for distributed large-scale data processing. Spark’s Application Programming Interface (API) has a powerful easy-to-use distributed abstractions similarly related to functional programming (e.g., map, filter, reduce) in several different languages. However, writing an efficient Spark applications is still error-prone, time-consuming, and requires a clear and deep understanding of the inner-workings of Spark. For instance, the same task can be implemented in several different ways, yet the execution time can vary drastically between them. For this, we introduce TaBOS, a rewrite-based optimizer for Spark programs. TaBOS takes a Spark job and automatically generates a state-space of equivalent optimized jobs using a set of semantics-preserving rewrite rules. Then, from the generated state-space, it selects one optimal program based on a predefined strategy. We introduce several selection strategies (e.g., job with maximum number of applied rewrite rules, job with minimum number of heavy operations) for identifying an optimal job from the generated state-space. We evaluate the effectiveness, robustness and speedup gain of our solutions using several case studies.

Comparative Analysis of Energy-Efficient Scheduling Algorithms for Big Data Applications

Nowadays, big data analytics has been widely applied in addressing the growing cybercrime threats. However, energy consumption is explosive increasing with the fast growth of big data processing in anti-cybercrime. In this paper, an energy-efficient framework for big data applications is proposed to reduce energy consumption while satisfying deadline constrains. First, the problem of energy-efficient tasks scheduling of a single Spark job is modeled as an integer program. We design an energy-efficient tasks scheduling algorithm to minimize the energy consumption for big data application in Spark. To avoid service-level agreement violations for execution time, we propose an optimal task scheduling algorithm with deadline constrains by tradingoff execution time and energy consumption. Experiments on a Spark cluster are performed to determine the energy consumption and execution time for several workloads from the HiBench benchmark suite. Our algorithms consume less energy on average than FIFO and FAIR under deadlines. The optimal algorithm is able to find near optimal tasks schedules to trade off energy consumed and response time benefit in small shuffle partitions.

Adding data provenance support to Apache Spark.

Debugging data processing logic in data-intensive scalable computing (DISC) systems is a difficult and time-consuming effort. Today's DISC systems offer very little tooling for debugging programs, and as a result, programmers spend countless hours collecting evidence (e.g., from log files) and performing trial-and-error debugging. To aid this effort, we built Titian, a library that enables data provenance-tracking data through transformations-in Apache Spark. Data scientists using the Titian Spark extension will be able to quickly identify the input data at the root cause of a potential bug or outlier result. Titian is built directly into the Spark platform and offers data provenance support at interactive speeds-orders of magnitude faster than alternative solutions-while minimally impacting Spark job performance; observed overheads for capturing data lineage rarely exceed 30% above the baseline job execution time.

Создание виртуальных кластеров Apache Spark в облачных средах с использованием систем оркестрации

Apache Spark is a framework providing fast computations on Big Data using MapReduce model. With cloud environments Big Data processing becomes more flexible since they allow to create virtual clusters on-demand. One of the most powerful open-source cloud environments is Openstack. The main goal of this project is to provide an ability to create virtual clusters with Apache Spark and other Big Data tools in Openstack. There exist three approaches to do it. The first one is to use Openstack REST APIs to create instances and then deploy the environment. This approach is used by Apache Spark core team to create clusters in propriatary Amazon EC2 cloud. Almost the same method has been implemented for Openstack environments. Although since Openstack API changes frequently this solution is deprecated since Kilo release. The second approach is to integrate virtual clusters creation as a built-in service for Openstack. ISP RAS has provided several patches implementing universal Spark Job engine for Openstack Sahara and Openstack Swift integration with Apache Spark as a drop-in replacement for Apache Hadoop. This approach allows to use Spark clusters as a service in PaaS service model. Since Openstack releases are less frequent than Apache Spark this approach may be not convenient for developers using the latest releases. The third solution implemented uses Ansible for orchestration purposes. We implement the solution in loosely coupled way and provide an ability to add any auxiliary tool or even to use another cloud environment. Also, it provides an ability to choose any Apache Spark and Apache Hadoop versions to deploy in virtual clusters. All the listed approaches are available under Apache 2.0 license.

Titian

Debugging data processing logic in Data-Intensive Scalable Computing (DISC) systems is a difficult and time consuming effort. Today's DISC systems offer very little tooling for debugging programs, and as a result programmers spend countless hours collecting evidence (e.g., from log files) and performing trial and error debugging. To aid this effort, we built Titian, a library that enables data provenance-tracking data through transformations-in Apache Spark. Data scientists using the Titian Spark extension will be able to quickly identify the input data at the root cause of a potential bug or outlier result. Titian is built directly into the Spark platform and offers data provenance support at interactive speeds-orders-of-magnitude faster than alternative solutions-while minimally impacting Spark job performance; observed overheads for capturing data lineage rarely exceed 30% above the baseline job execution time.

Complement or competition: Latino employment in a nontraditional settlement area

The migration of Latinos to nontraditional settlement areas in the United States is renewing interest in how an established low-skilled work force is affected by the inflow of a minority group whose members tend to have a weak basket of human capital. Some scholars focus on how the incoming group creates head-to-head competition with established workers. An alternative view posits that, depending on the context of the receiving labor market, incoming workers may primarily fill roles that complement preexisting labor market arrangements. I study these issues in the region of the country that has experienced the most pronounced in-migration of Latinos during the past few years. The findings indicate migrating Latinos tend to complement preexisting labor market conditions rather than spark job competition and undercut the earning power of non-Latinos.