Building block components to control a data rate in the Apache Hadoop compute platform

Abstract

Abstract-Resource management is one of the most indispensable components of cluster-level infrastructure layers. Users of such systems should be able to specify their job requirements as a configuration parameter (CPU, RAM, disk I/O, network I/O) and have the scheduler translate those into an appropriate reservation and allocation of resources. YARN is an emerging resource management in the Hadoop ecosystem, which supports only RAM and CPU reservation at present.In this paper, we propose a solution that takes into account the operation of the Hadoop Distributed File System to control the data rate of applications in the framework of a Hadoop compute platform. We utilize the property that a data pipe between a container and a DataNode consists of a disk I/O subpipe and a TCP/IP subpipe. We have implemented building block software components to control the data rate of data pipes between containers and DataNodes and provide a proof-of-concept with measurement results.

Provisioning Input and Output Data Rates in Data Processing Frameworks

“…Motivated by the need, we design a complete solution that can be applied to provision the I/O data rate of applications in both the Mesos and YARN frameworks. Note that this is a result 1 that has gradually been improved over the years based on our previous experiences [10,11,24]. We demonstrate that the proposed functionalities can be integrated into two popular data processing frameworks such as Mesos and YARN to control the I/O data rates (disk I/O and network I/O) of applications, which may relieve the pain of service providers on the integration of schedulers to existing frameworks.…”

“…The idea of co-locating different scheduler frameworks in the same data center has been in discussion for the benefit of the operators and service providers. For example, only HDFS read traffic shaping in a YARN cluster using Traffic Control (LTC) mechanism was proposed in [11], while the disk I/O problem for Hadoop MapReduce applications and Spark applications [3,28] was investigated in [10,24]. Xu and Zhao [27] presented an interposed big-data I/O scheduler to provide I/O performance differentiation for competing applications in a shared big-data system.…”

“…Since a computing cluster can be built up of heterogeneous hardware and software components, the I/O data rate perceived by applications is unpredictable due to the contention for the resources of the physical servers, and there is a need to monitor applications running on these platforms as well [12]. As shown in [11,24], the I/O contention of applications leads to the degradation of quality of service.…”

BME VIK Annual Research Report on Electrical Engineering and Computer Science 2015

“…Some customers may require a data rate guarantee because their jobs should be finished by a certain deadline. Therefore, the provision of the quality of service regarding a data rate guarantee may play a key factor to attract customers [69,70].…”

“…In recent collaboration works with Nokia [69,70], we proposed a set of functionality to monitor and isolate I/O demands in production environments. The proposed functionality can be used to minimize contention situations that lead to the I/O degradation offered to applications and clients.…”