RMBench: A Benchmarking Suite for Distributed Real-Time Middleware

Bibliographic Details
Main Author: Delaney, Matthew
Language:English
Published: Ohio University / OhioLINK 2005
Subjects:
Online Access:http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1125371173
id ndltd-OhioLink-oai-etd.ohiolink.edu-ohiou1125371173
record_format oai_dc
spelling ndltd-OhioLink-oai-etd.ohiolink.edu-ohiou11253711732021-08-03T05:43:08Z RMBench: A Benchmarking Suite for Distributed Real-Time Middleware Delaney, Matthew Computer Science RMBench Real-Time Benchmarking Distributed real-time Distributed Middleware Quality of service <p>Numerous middleware packages for resource management and communications are emerging, however there is no toolkit available to systematically evaluate and compare them. In [8], Sharp performs experimental assessment of real-time Java middleware, and provides a list of important metrics that are useful in such evaluations. Tools to automate collecting of such metrics have been developed, but previous work in distributed real-time benchmarking has not fully addressed the problem domain.</p> <p>The Hartstone Distributed Benchmark (HDB) [2] [1] allows periodic and aperiodic work to be increased gradually, in order to find the limitations of hardware, operating systems and middleware. Dynbench [22] provides periodic, event-driven, and event-driven periodic workloads by simulating an air defense system. The Synthetic Workload Specification Language (SWSL) [14] provides a formal specification language for synthetic workloads in which real-time constraints can be specified, and Synthetic Workload Generator (SWG) [15] is used to implement SWSL specifications.</p> <p>HDB, Dynbench, and SWSL/SWG are all useful for finding the performance limits of real-time infrastructure components, however they do not support arbitrary workloads or QoS parameters. These deficiencies have been addressed in the RMBench system, a set of CORBA services for benchmarking the performance of distributed real-time middleware.</p> <p>Major features of RMBench include Sensor and Path Nodes. Users can create Sensor processes that periodically distribute dynamic workloads. These Nodes can simulate scenarios where the workloads in the system change during runtime by adjusting the amount of data coming from a Sensor. Specifically, a CORBA service is provided for changing the periods of Sensor processes and for changing the amount of outgoing data (workload) generated by Sensor processes. Users can also create Path Node processes, which receive messages, perform work, and pass on messages to other Path Node processes. To emulate dynamic real-time systems, the execution time of a Path Node can be defined to be a function of the node’s workload. Mathematical workload functions such as log n, n log n, and n<sup>m</sup> (where n is the size of the current message) can be defined and changed during run-time. A CORBA service allows resource management middleware to dynamically change Quality of Service (QoS) parameters of Path Node processes.</p> <p>Other features of RMBench include an Experiment Specification File which is used to create automated, repeatable experiments. A customizable display process which permits viewing of end-to-end performance, node performance, and workloads. And finally, Experiment Log Files are provided to facilitate offline analysis.</p> <p>RMBench has been used to validate an integration of the QARMA [9] resource management software and a distributed scheduling service. Extensive experiments are run and the benefit of this integration is show by using RMBench.</p> 2005-10-10 English text Ohio University / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1125371173 http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1125371173 unrestricted This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.
collection NDLTD
language English
sources NDLTD
topic Computer Science
RMBench
Real-Time
Benchmarking
Distributed real-time
Distributed Middleware
Quality of service
spellingShingle Computer Science
RMBench
Real-Time
Benchmarking
Distributed real-time
Distributed Middleware
Quality of service
Delaney, Matthew
RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
author Delaney, Matthew
author_facet Delaney, Matthew
author_sort Delaney, Matthew
title RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
title_short RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
title_full RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
title_fullStr RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
title_full_unstemmed RMBench: A Benchmarking Suite for Distributed Real-Time Middleware
title_sort rmbench: a benchmarking suite for distributed real-time middleware
publisher Ohio University / OhioLINK
publishDate 2005
url http://rave.ohiolink.edu/etdc/view?acc_num=ohiou1125371173
work_keys_str_mv AT delaneymatthew rmbenchabenchmarkingsuitefordistributedrealtimemiddleware
_version_ 1719424106919624704