Abstract
We consider efficient network provisioning algorithms
for applications that aggregate large data files from multiple remote sites
to a central facility (where the aggregated data is further processed). Many
important bandwidth-hungry scientific applications use such data aggregation,
and it is important to efficiently use network resources to meet their requirements.
We term an entire large-scale data-aggregation session as a data-aggregation
request (DAR). In this paper, we investigate the problem of on-demand provisioning
of DARs over a wavelength-division multiplexing (WDM) backbone network infrastructure.
Our DAR provisioning problem is challenging, as for each DAR we need to jointly
identify lightpaths (route, assign wavelengths, and groom) for each of the
to-be-transferred files, and schedule DAR's file transfers in time. We first
model our DAR provisioning problem mathematically as a mixed integer linear
program (MILP); to solve our problem in practice, we propose a DAR provisioning
algorithm (named DARP). From our numerical results, we find DARP to be efficient
when compared with other benchmark algorithms. We study DARP's performance
for a varying number of aggregating sites deployed in the network (i.e., sites
with supercomputer facilities) and perform a detailed sensitivity analysis
on several parameters of our problem. We also investigate the effect of partitioning
the data to be transferred into pieces and conclude that, if the partitioning
method is carefully designed, slight improvement over the approach that transfers
the whole file (DARP) is possible.
© 2009 IEEE
PDF Article
More Like This
Cited By
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
Contact your librarian or system administrator
or
Login to access Optica Member Subscription