China
Africa
Explore chapters and articles related to this topic
Resource Management for MapReduce Jobs Performing Big Data Analytics
Published in Kuan-Ching Li, Hai Jiang, Albert Y. Zomaya, Big Data Management and Processing, 2017
Norman Lim, Shikharesh Majumdar
First, MRBB-RM is compared with the Minimum Resource Quota Earliest Deadline First with Work-Conserving Scheduling (MinEDF-WC) technique [7], which has objectives that are similar to that of MRBB-RM. MinEDF-WC is a resource allocation and scheduling technique for processing MapReduce jobs with deadlines that is based on the well-known earliest deadline first scheduling policy. MinEDF-WC allocates the minimum number of resources required for completing a job before its deadline and also has the ability to dynamically allocate and deallocate resources from active jobs when required. This ability to dynamically allocate and deallocate resources allows a machine with spare resources to share its unused resources with other jobs that need them. A comparison between the performance of MRBB-RM with that of MinEDF-WC is presented next. The workload used is a synthetic workload generated from workload traces of a Hadoop cluster used at Facebook in October 2009 that is described in Reference 7.
Broadband Networks
Published in Naoaki Yamanaka, High-Performance Backbone Network Technology, 2020
Priority queues have limited applicability if it is necessary to identify more than two or three service classes or if it is necessary to satisfy individually specified flow quality of service requirements. Much more complex mechanisms, including “weighted fair queueing” [16] and “earliest deadline first” scheduling [8], have been extensively studied in the last few years. There are two main motivations: to satisfy deterministic delay guarantees;to protect individual flows from the traffic of other users.
Real-Time Operating Systems
Published in Leanna Rierson, Developing Safety-Critical Software, 2017
Earliest deadline first scheduling. This is a dynamic priority preemptive policy. It places tasks in a priority queue, so that whenever a task finishes, the queue is searched for the process closest to its deadline.* This process becomes the next task scheduled for execution. Basically, the scheduler selects the process that has the earliest deadline to run first, which preempts any processes with a later deadline.
A real-time logo detection system using data offloading on mobile devices
Published in Cyber-Physical Systems, 2018
At the same time, we want to reduce the bandwidth usage. Ideally, every frame should be sent to the server for further processing and detection. However, this will waste bandwidth and server resources since some frames may not have a logo or multiple frames may have the same logo. We need to determine locally which frame is important so that mobile devices can send only important information to the server. It is also difficult to determine the important frames locally since running new logo detection on mobile devices usually needs at least of hundreds of milliseconds, which is not practical in real AR applications. In our system, we propose a local model to determine which frame is important by monitoring view change events. The key observation is that the camera often captures new logos when user’s view changes. After offloading detection jobs to the back-end server, a proper scheduling strategy should be designed to provide good performance for most connected users. In our system, we adopt the earliest deadline first scheduling strategy to minimise the maximum latency for all users.