Our Senior Correspondent
THIRUVANANTHAPURAM, Jan. 8
The computing power available on homogeneous networks can now be
put to better use, thanks to a newly-designed parallel
programming environment, Parallel and Distributed Computing
(Paradisc), which helps in the execution of the user's tasks and
provides access to resources beyond that available through a
standard system.
A paper, titled `Paradisc: A cost-effective model for parallel
and distributed computing', was presented at the third
international conference on `High-performance computing (HiPC)'
hosted by the Technopark here.
A team of experts from the Centre for Development of Advanced
Computing (CDAC), Bangalore, said in the paper that Paradisc, as
a scalable openframe computing model, provides the transparency
needed in a network of homogeneous system of PCs, workstations
and minicomputers connected through a local area network (LAN) or
a wide area network (WAN).
The team comprised Mr. R. Achutha Raman, Mr. Rajkumar, Mr. G.
Hari Prakash and Mr. B. Bala Kishore of the CDAC. According to
the experts, Paradisc supports three models of computing by
providing the functions required to view any homogeneous network
as a loosely coupled parallel computer, processor pool
architecture or a cluster of workstations.
It aims at providing a cost-effective parallel and distributed
programming environment to academic and R&D institutions as it
employs the existing LAN, and models it to support both the
paradigms.
By relying solely on commodity hardware and software, a network
of workstations can offer parallel processing at a low cost. Such
a network can be realised as a processor bank in which dedicated
processors provide computing cycles. It can also consist of a
dynamically varying set of machines that perform long-running
computations during idle periods.
Paradisc provides a software model of integrated solutions not
found in a single unit in the earlier systems and precludes any
change in the underlying current set-up, including the network
and operating system. The main aim is to build a transparent,
distributed and parallel environment which is flexible, cost-
effective and simple too.
Paradisc, therefore, is a software model which provides the best
of both worlds, parallel and distributed, the experts claimed. It
uses the existing network of homogeneous computers to run
parallel programmes using the logical connectivity method and
executes distributed programmes utilising the power of other idle
computers and pool processors in the network. The basic idea is
to provide users with the illusion of a single but powerful time-
sharing system.
Paradisc is typified by specialised components such as the
Paradisc server, which includes a configuration server, a
processor pool server and a resource server. The Paradisc server
configures a group of computers logically to the required
topology and it allocates idle machines and pool processors for
users on demand.
As a loosely coupled parallel machine, the network can be
configured to standard topologies or user-defined topologies
suitable to the problem domain. These topologies are logically
connected through the configuration server provided by Paradisc.
The network can be configured into any of the two modes, parallel
or distributed, or both, with the use of two different
configuration files. Each node maintains a per-process node
connectivity status. It enables the identification of
neighbourhood nodes. The configuration server maintains the
global per-process connectivity information. The interprocess
communication is provided by a process communication interface.