Information Technology Reference
In-Depth Information
2.3.7
Grid-Enabled Instruments and Sensor Networks
Technological advances over the last several decades have greatly increased
our ability to acquire massive quantities of data via various scientii c instru-
ments and sensors. At the same time, however, models, paradigms, and
middleware for actually utilizing these data have not kept pace. Research in
data curation, provenance, and metadata are crucial for scientists to exploit
these large new sources of data. As the source of data, instruments and sen-
sors are a crucial part of the data management process. Previously, sensors
and instruments could be treated as off-grid entities, because the volume
and rate of data were low. Such systems are no longer adequate, however.
Along with collaborators at Indiana University, the University of
Wisconsin, and the San Diego Supercomputer Center, we are developing
cyberinfrastructure to more fully integrate instruments and sensors into
the grid. We are developing systems for remote X-ray crystallography to
allow valuable instruments to be better utilized (Bramley et al., 2006;
Devadithya et al., 2005; McMullen et al., 2005). We are also working on
systems to better capture data provenance and metadata from sensor sys-
tems (Pan et al., 2006; Skovronski and Chiu, 2006).
2.3.8
Grid Emulation Framework for Multicore Processors
The microprocessor industry is rapidly moving toward chip multiproces-
sors (CMPs), commonly referred to as multicore processors, where multi-
ple cores can independently execute different threads. This change in
computer architecture requires corresponding design modii cations to
programming paradigms, including grid middleware tools, in order to
harness the opportunities presented by multicore processors. Naive
implementations of grid middleware on multicore systems can severely
impact performance due to limitations of shared bus bandwidth, cache
size and coherency, and communication between threads. The goal of
developing an optimized multithreaded grid middleware for emerging
multicore processors will be realized only if researchers and developers
have access to an in-depth analysis of the impact of several low-level
microarchitectural parameters on performance. None of the current grid
simulators and emulators provide feedback at the microarchitectural level,
which is essential for such an analysis. We have designed and developed
a prototype emulation framework, the Multi-core Grid (McGrid), to
analyze and provide insightful feedback on the performance limitations,
bottlenecks, and optimization opportunities for grid middleware on mul-
ticore systems. McGrid is designed as a highly coni gurable framework
with the ability to provide feedback at various levels of granularity, and
for different hardware types and coni gurations used in heterogeneous
grid environments. Using the McGrid framework, we can study the effect
of various design options on the efi ciency of key microarchitectural
 
Search WWH ::




Custom Search