Adaptive data parallel computing on workstation clusters
نویسندگان
چکیده
Many real world problems can be categorised under the data parallel programming paradigm. Such problems can be solved in the workstation cluster environment, which is fast becoming a viable platform for parallel computing, by allocating each node a portion of data proportional to its computational capability. In the workstation cluster environment, workstations may be recalled by their primary user(s), while at times the workstation becomes so heavily loaded with work given by its primary user(s) that it becomes a bottleneck for performance of the parallel job. In such situations, the need for deleting a node arises, and the partitions need to be readjusted to reeect the new connguration of the virtual parallel machine. Conversely , nodes may become available to join the parallel machine reaped from the networked environment. In this paper policies for readjusting data partitions following addition/deletion of nodes are presented. Results of preliminary investigations show the eeectiveness of these policies.
منابع مشابه
Runtime Support for Replicated Parallel Simulators of an ATM Network on Workstation Clusters
An eeective approach of speeding up the simulation of an ATM network on workstation clusters is presented. In this approach, multiple simulation runs are performed by replicated parallel simulators (RPSs) concurrently. Since the execution platform of the simulation is in a shared-network environment, the RPSs must compete with other applications for resources. The RPSs support adaptive executio...
متن کاملApproaches to Support Parallel Programming on Workstation Clusters: A Survey
The goal of this report is to survey state of the art and existing approaches for parallel programming on workstation clusters with special emphasis on object{oriented programming. First, workstation clusters as parallel computing platforms are characterized and fundamental concepts for parallel programming are discussed. Then, an overview of existing tools, systems, languages, and environments...
متن کاملExperiences with Asynchronous Communication Models in VEOS, a Distributed Programming Facility for Uniprocessor LANs
Like conventional multiprocessors, workstation clusters can provide data sharing and parallel computing. But unlike multiprocessors, these clusters provide flexible connectivity and can tolerate heterogeneous processing elements. Uniprocessor LANs are a common choice for cost-effective computing. The workstation nodes typically run a version of Unix and support common Unix services such as reli...
متن کاملLoad Balancing for Parallel Loops in Workstation Clusters
Load imbalance is a serious impediment to achieving good performance in parallel processing. Global load balancing schemes cannot adequately manage to balance parallel tasks generated from a single application. Dynamic loop scheduling methods are known to be useful in balancing parallel loops on shared-memory multiprocessor machines. However, their centralized nature causes a bottleneck even fo...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
- J. Parallel Distrib. Comput.
دوره 64 شماره
صفحات -
تاریخ انتشار 2004