November 20, 2006
eXludus Technologies Inc. introduced version 2.1 of RepliCator, the company's software product for compute clusters and grids.
"RepliCator 2.1 adds two important capabilities that are world firsts," said eXludus Technologies CEO Benoît Marchand. "Our new features, Asynchronous Results Transfer (ART) and Groups, expand the range of applications for which RepliCator can substantially boost performance. ART can make an important difference in the manufacturing, life sciences, financial services, EDA and public sector research markets, as well as in high performance computing wherever checkpoints or output data sets are bottlenecks today."
According to eXludus, RepliCator 2.1 is the world's first cluster/Grid software to provide fault-tolerant and data resilient data transfer from any source to any destination. RepliCator's Asynchronous Results Transfer (ART) capability can be applied to output files (e.g., intermediate results, checkpoints, final results) or to input files.
On a real-world workload using checkpoint files, Durham University's Institute for Computational Cosmology (UK) improved cluster throughput by a factor of 1.48 by using RepliCator to allow overlapping of output checkpoint data movement with computation. Wall clock time was reduced from 445 seconds to 301 seconds, and I/O wait time shrank from 145 seconds to 0.1 seconds
"It doesn't matter what the underlying application is. If it involves large files and is doing substantial checkpointing on a medium to large-size cluster, RepliCator's new ART feature can make a major difference," said Marchand.
With ART, data can be transferred at the end of a task automatically, using eXludus's MetaLanguage Processor, or manually through a command-line interface, or on-the-fly through the API interface. RepliCator allows the compute nodes to get back to what they do best -- run your simulation.
RepliCator 2.1 also offers the ability for parallel file serving of an input file to a targeted subset of cluster nodes, as opposed to the entire cluster. With RepliCator 2.1's new Groups feature, the cluster subset can be made to match the workload manager's group definition (e.g., "Send this Nastran input file to Nastran nodes"). The Groups feature allows an unlimited number of groups to be used concurrently on a cluster.
"Groups allows RepliCator to operate in multi-application clusters more efficiently, by caching data only on likely target nodes, those that are candidates to run certain job types," according to Marchand.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?