Global Positioning System Reference
In-Depth Information
GPU-based queries have turned out to be limited only by bus bandwidth,
not by operational complexity, for a large class of queries (Baumann et al.
2009).
Oracle Spatial GeoRaster makes use of the parallel processing
framework available in the Oracle DBMS, but further implements
parallelization specifi c to raster processing operations, since Oracle's
generic framework itself is optimized for row-based data (Xie et al. 2012).
Parallelization of raster processing starts by splitting the raster object into
several subsets depending on the degree of parallelism, or user input. Then
the Oracle parallel execution framework splits the task into subtasks that
will each process one subset independently of the others, and distributes
them to separate subprocesses for execution. Benchmarks have shown that
this results in signifi cant performance speedup in comparison to a non-
parallelized Oracle implementation. This approach resembles tile-based/
chunk-based processing similar to the other architectures presented, except
for the additional overhead of managing heavy-weight operating system
processes. It should be noted, though, that Oracle GeoRaster does not come
with a general-purpose raster query language but supports only sequential
execution of predefi ned functions.
Cloud processing
Cloud computing is an internet based computing model that enables easy
and dynamic provisioning, confi guring, and de-provisioning of a pool of
resources (such as compute, storage, data, network, applications, services)
that are used as needed to satisfy a workload requirement. According to the
US National Institute of Standards and Technology (NIST), the fi ve main
characteristics of cloud computing include (Mell and Grance 2011):
on-demand self-service:
resource provisioning by the user without human
interaction at the server;
broad network access:
resource accessibility to different devices from
anywhere over the network;
resource pooling:
resources are shared among multiple consumers in a
non-dedicated manner;
rapid elasticity:
scaling out/in of resources in proportion commensurate
with the user's demand;
measured service:
metering and measuring of resource usage.
Similarly, NIST defi nes cloud deployment models which can follow
either a private (used within a single organization), a community (used
within a community of consumers with shared concerns), a public (use
open to the public), or a hybrid cloud computing model (any combination
of the aforementioned variants).
Search WWH ::
Custom Search