An empirical analysis of scheduling techniques for real-time cloud-based data processing

Abstract

In this paper, we explore the challenges and needs of current cloud infrastructures, to better support cloudbased data-intensive applications that are not only latency-sensitive but also require strong timing guarantees. These applications have strict deadlines (e.g., to perform time-dependent mission critical tasks or to complete real-time control decisions using a human-in-the-loop), and deadline misses are undesirable. To highlight the challenges in this space, we provide a case study of the online scheduling of MapReduce jobs executed by Hadoop. Our evaluations on Amazon EC2 show that the existing Hadoop scheduler is ill-equipped to handle jobs with deadlines. However, by adapting existing multiprocessor scheduling techniques for the cloud environment, we observe significant performance improvements in minimizing missed deadlines and tardiness. Based on our case study, we discuss a range of challenges in this domain posed by virtualization and scale, and propose our research agenda centered around the application of advanced real-time scheduling techniques in the cloud environment.Abstract-In this paper, we explore the challenges and needs of current cloud infrastructures, to better support cloud-based data-intensive applications that are not only latency-sensitive but also require strong timing guarantees. These applications have strict deadlines (e.g., to perform time-dependent mission critical tasks or to complete real-time control decisions using a human-in-the-loop), and deadline misses are undesirable. To highlight the challenges in this space, we provide a case study of the online scheduling of MapReduce jobs executed by Hadoop. Our evaluations on Amazon EC2 show that the existing Hadoop scheduler is ill-equipped to handle jobs with deadlines. However, by adapting existing multiprocessor scheduling techniques for the cloud environment, we observe significant performance improvements in minimizing missed deadlines and tardiness. Based on our case study, we discuss a range of challenges in this domain posed by virtualization and scale, and propose our research agenda centered around the application of advanced real-time scheduling techniques in the cloud environment.

Real-Time Information Derivation from Big Sensor Data via Edge Computing

“…The problem of meeting real-time deadlines in Hadoop is investigated in [11][12][13][14][15][16]. Also, in [17], admission control is applied to deadline-driven batch data analysis tasks.…”

“…It is complementary to our work in that EDF adopted in this paper is a dynamic priority scheduling method. Moreover, we support cost-effective transfer and streaming of sensor data from IoT devices to the edge server on which RTMR runs as well as memory reservation and pipelining of intermediate data between map-reduce phases different from [11][12][13][14][15][16][17]47]. Phoenix [20] supports efficient in-memory execution of map/reduce tasks in a multicore system.…”

“…In this paper, we extend the notion of in-memory map-reduce via real-time scheduling, memory reservation, and data streaming and pipelining to support real-time data analytics, while supporting adaptive sensor data transfer based on data importance. Because there is no standard real-time big data framework and different approaches (e.g., [11][12][13][14][15][16][17]47]) use different frameworks, deadlines, and scheduling algorithms, it is very hard (if at all possible) for us to implement and empirically compare all related approaches to RTMR on our own. Instead, we compare the performance of RTMR to that of Phoenix [20] that is most closely related to our work as a proof of concept, while qualitatively comparing our approach to advanced related methodologies [11][12][13][14][15][16][17]47] as discussed above.…”

Optimal Map Reduce Job Capacity Allocation in Cloud Systems

“…Deadlines for MapReduce jobs are considered also in [23]. The authors recognize the inability of Hadoop schedulers to handle properly jobs with deadlines and propose to adapt to the problem some well-known multiprocessor scheduling policies.…”

Assessing the Impact of Concurrent Replication with Canceling in Parallel Jobs

“…While this approach improves the job reliability, it generates additional delays that degrade the quality of service (QoS) offered by the application, a major issue for latency-sensitive applications [2]. An alternative is to process the replicas concurrently, improving the job reliability and limiting the latency.…”