Impact Of Load Balancing On Unstructured Adaptive Grid Computations For Distributed-Memory Multiprocessors
نویسندگان
چکیده
The computational requirements for an adaptive solution of unsteady problems change as the simulation progresses. This causes workload imbalance among processors on a parallel machine which, in turn, requires significant data movement at runtime. We present a new dynamic loadbalancing framework, called JOVE, that balances the workload across all processors with a global view. Whenever the computational mesh is adapted, JOVE is activated to eliminate the load imbalance. JOVE has been implemented on an IBM SP2 distributed-memory machine in MPI for portability. Experimental results for two model meshes demonstrate that mesh adaption with load balancing gives more than a sixfold improvement over one without load balancing. We also show that JOVE gives a 24-fold speedup on 64 processors compared to sequential execution.
منابع مشابه
Load Balancing Unstructured Adaptive Grids for CFD Problems
Mesh adapt(on is a powerful tool for efficient unstructured-grid computations but causes load imbalance among processors on a parallel machine. A dynamic load balancing method is presented that balances the workload across all processors with a global view. After each parallel tetrahedral mesh adapt(on, the method first determines if the new mesh is sufficiently unbalanced to warrant a repartit...
متن کاملExperiments with Repartitioning and Load Balancing Adaptive Meshes
Mesh adaption is a powerful tool for efilcient unstructured-grid computations but causes load imbalance on multiprocessor systems. To address this problem, we have developed PLU M, an automatic portable framework for performing adaptive largescale numerical computations in a message-passing environment. This paper presents several experimental results that verify the effectiveness of PLUM on se...
متن کاملEnhanced Dynamic Load-Balancing of Adaptive Unstructured Meshes
Modern PDE solvers increasingly use adaptive unstructured meshes in order to discretise complex geometries and control numerical error The problem of dividing up the domain equally for a distributed memory parallel computer whilst minimising the inter subdomain dependencies can be tackled with graph based methods such as Recursive Spectral Bisection This paper describes an extension to such met...
متن کاملDynamic load-balancing for PDE solvers on adaptive unstructured meshes
Modern PDE solvers written for time-dependent problems increasingly employ adaptive unstructured meshes (see Flaherty et al. [4]) in order to both increase e ciency and control the numerical error. If a distributed memory parallel computer is to be used, there arises the signi cant problem of dividing up the domain equally amongst the processors whilst minimising the intersubdomain dependencies...
متن کاملUsing Spanning-Trees for Balancing Dynamic Load on Multiprocessors - Distributed Memory Computing Conference, 1991. Proceedings., The Sixth
We consider the problem of load balancing t o minimize the cost of dynamic computations, including the cost of migrations. We analyze the costs associated with diffusion based algorithms for several common architectures. We introduce the Ripple load balancing paradigm, which has several advantages over diffusion methods, including flexibility and faster convergence.
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:
دوره شماره
صفحات -
تاریخ انتشار 1996