Improved Parallel I/O via a
                  Two-phase Run-time Access Strategy* 

             (Presented in: IPPS `93 Parallel I/O Workshop)


      Juan Miguel del Rosario 	Rajesh Bordawekar	Alok Choudhary

                 Northeast Parallel Architectures Center
                      111 College Place, Room 3-201
                          Syracuse University
                        Syracuse, NY 13244-4100


                            February 9, 1993


                                ABSTRACT 

As scientists expand their models to describe physical phenomena of 
increasingly large extent, the memory capacity of parallel machines become 
insufficient to contain all the required computational data, and I/O becomes 
crucial. Thus, a system with limited I/O capacity can severely constrain the 
performance of the entire program. This problem has become critical enough 
that most parallel computers now provide some measure of support for parallel 
I/O.

We provide experimental results, performed on an Intel Touchtone Delta and 
nCUBE 2 I/O system, to show that the performance of existing parallel I/O 
systems can vary by several orders of magnitude as a function of the data 
access pattern of the parallel program. We then propose a two-phase access 
strategy, to be implemented in a runtime system, in which the data 
distribution on computational nodes is decoupled from storage distribution. 
In the first phase, data is accessed based on the conforming distribution 
from the I/O system; subsequently, in phase 2, it is redistributed at 
run-time. Our experimental results show that performance improvements of 
several orders of magnitude over direct access based data distribution 
methods can be obtained, and that performance for most data access patterns 
can be improved to within a factor of 2 of the best performance. Further,the 
cost of redistribution is a very small fraction of the overall access cost. 

Juan Miguel del Rosario
CIS Dept.
111 College Place
Syracuse University
Syracuse, NY 13244-4100
mrosario@sunlab.syr.edu
(315) 443-1722

Rajesh Bordawekar
ECE Dept.
121 Link Hall
Syracuse University
Syracuse, NY 13244-4100

Alok Choudhary
ECE Dept.
121 Link Hall
Syracuse University
Syracuse, NY 13244-4100
choudhar@cat.syr.edu
(315) 443-4280