DCSC logo
+Open all         -Close all
 The Horseshoe5 and Horseshoe6 supercomputers belong to the family of
 Beowulf Class II clusters. This type of clusters are based on off-the-shelf
 components with added specialized hardware to achieve extra performance.

 In our setup we use 1U standard servers interconnected with Infiniband.

 The system presents itself as a frontend machines which you can log in to.
 The frontend machine is only used to compile programs and stage runs on
 execution nodes. These execution nodes are under the control of a (computing)
 resource management system or batch queue system. PBS/TORQUE and MAUI is used
 for this purpose.
 From a practical point of view, The Horseshoe is simply some batch queue
 systems supported by a lot of execution hosts, which will allow many jobs to
 run at the same time. These jobs can be of scalar or parallel type.

 Since the Horseshoe is "simply" a network of individual workstations 
 parallel applications which rely on MP (OpenMP) libraries, or SMP types of
 shared memory cannot be run on the Horseshoe. Applications using MPI, PVM,
 or other types of network based interprocess communication are supported. 

 Please read more at these sites: