MGT 7-1.indd - KMI Media Group
MGT 7-1.indd - KMI Media Group
MGT 7-1.indd - KMI Media Group
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
to be spread across commodity hardware<br />
and software grids to enable its near realtime<br />
completion and delivery,” said Cuppan.<br />
“Cloud architecture concepts may allow us<br />
to address processing transients and the<br />
fluid plug-and-play objectives of coalition<br />
operations.”<br />
VIRTUALIZATION MIDDLEWARE<br />
Meanwhile, industry innovations and<br />
implementations plow ahead, not waiting<br />
for the standards. One example is the way<br />
in which GeoEye and Appistry are working<br />
together to cloud-enable key applications<br />
using Appistry’s Enterprise Application Fabric<br />
(EAF) to create a fault-tolerant cloud.<br />
“We’re using the Appistry EAF fabric to move<br />
to a distributed processing system,” Helmering<br />
said.<br />
GeoEye supplies the algorithms and key<br />
software, while Appistry provides the EAF<br />
software. “We built frameworks on top of<br />
EAF that are customized for geospatial use.<br />
GeoEye is using it in production now for<br />
sensor data coming from a variety of satellites,<br />
taking raw data and creating refined<br />
product,” said Lozano.<br />
“We’ve gone from prototyping to building<br />
real software,” he said. “We’re cloudenabling<br />
GeoEye’s product that massages<br />
incoming data so they can run their applications<br />
on private clouds to achieve higher<br />
scales more quickly at lower infrastructure<br />
and operational costs by using off-the-shelf<br />
hardware.”<br />
EAF is virtualization middleware that<br />
is highly scalable and aggregated, pulling<br />
together different views of hardware<br />
to appear as one. “A key to cloud-enabling<br />
applications is how to aggregate the underlying<br />
resources, whether physical or virtual,”<br />
Lozano noted.<br />
EAF achieves fault tolerance by automatically<br />
reassigning work in the event of a<br />
processing problem.<br />
Lozano said EAF can cloud-enable applications<br />
easily in some cases by simply adding<br />
application metadata to tell EAF what the<br />
target application is doing. “When the situation<br />
is more complex, it includes [application<br />
programming interface] calls that are<br />
independent of the cloud,” he said.<br />
Lozano believes a missing ingredient that<br />
is essential to the success of cloud computing<br />
is the ability to make applications feel native<br />
to the world of the cloud.<br />
Using EAF is one way to cloud-enable<br />
applications. Another is Hadoop, an open<br />
source software platform for writing applications<br />
that process massive amounts of data<br />
on large clusters of commodity hardware.<br />
Hadoop uses the Hadoop Distributed File<br />
Systems (HDFS) to implement MapReduce,<br />
which divides applications into many blocks<br />
of work, while HDFS creates multiple replications<br />
of data blocks to achieve reliability.<br />
HDFS then places replications of data on<br />
nodes in a cluster.<br />
NGA has studied Hadoop as a blueprint<br />
by which to take an OGC geospatial metadata<br />
standard called Catalog Web Service (CSW)<br />
and distribute it broadly. “CSW backend<br />
implementation details that enable clients to<br />
discover data and related processing services<br />
might be implemented using HDFS, which<br />
are central cloud processing constructs for<br />
Google and Yahoo. The CSW indices may be<br />
distributed across a broad computational lattice,”<br />
said Cuppan.<br />
Both Amazon and Google have technology<br />
similar to Hadoop to cloud-enable<br />
applications, but it is kept proprietary as a<br />
way to encourage customers to bring their<br />
data to Amazon or Google’s cloud computing<br />
platform. Google’s Big Table is a compressed,<br />
high-performance RDMS built partly on<br />
the Google File System, which Google App<br />
Engine customers can access.<br />
However fast and petabyte-scale the<br />
Google Big Table database is, developers<br />
could need to rewrite their Big Table applications.<br />
“If you write your application to make<br />
use of Big Table, it’s proprietary and you can’t<br />
redeploy the application extensively without<br />
rewriting it,” Lozano said, adding that<br />
Appistry is working on a standard approach<br />
to minimize source code changes made in<br />
order to cloud-enable an application.<br />
FILE MANAGEMENT<br />
File management and security are essential<br />
issues with large global geospatial files.<br />
To ease existing file management issues<br />
resulting from complex, multiple network<br />
attached storage (NAS) or storage area network<br />
(SAN) systems that might not scale<br />
well, it is usually necessary to integrate storage<br />
and file systems.<br />
Establishing a common file system that<br />
is shared by Windows, Linux and Unix users<br />
can achieve seamless interoperability that<br />
carves a path to cloud computing. “In Windows,<br />
you would see a shared drive, like a D<br />
drive, that looks the same to one as another<br />
regardless of the underlying platform,” said<br />
IBM’s Ames.<br />
In one example, IBM’s existing, tested<br />
technology, the Scale Out File System<br />
(SOFS), accomplishes this objective. The<br />
SOFS is an NAS/grid system that manages<br />
and broadly scales out NAS environments,<br />
in part by optimizing storage. SOFS utilizes<br />
the company’s General Parallel File System,<br />
a highly secure, long-tested and established<br />
high performance computing system.<br />
IBM middleware, the Websphere Federation<br />
Server, contains an embedded RDMS<br />
and integrates remote, diverse data and content<br />
sources, making them appear as if<br />
they were the same database. The technology<br />
contains security management and user<br />
authentication. The SAIC Common Criteria<br />
Testing Lab successfully tested the security<br />
of Websphere Federation Server running on<br />
AIX and Red Hat Linux.<br />
“Using Websphere Federation Server, a<br />
user writes a single query, WRS then optimizes<br />
the query, sends the request across<br />
the heterogeneous information stores, which<br />
then joins the necessary information and<br />
sends it back to the user,” explained Ames.<br />
There is no question that the desire to<br />
achieve improved exploitation of available<br />
intelligence is leading to consolidated cloud<br />
strategies for geospatial information.<br />
“We have all these intelligence capabilities<br />
across the globe. It makes sense to integrate<br />
by cross-tipping or cross information<br />
providing so analysts could discover other<br />
capabilities. One example is combining imagery<br />
intelligence with signals intelligence. If<br />
you have a fixed high resolution image from<br />
a given time and a moving, lower resolution<br />
image, then it’s stronger to have the fixed<br />
image plus the moving image to acquire a<br />
much richer intelligence view,” said Ames.<br />
“By putting the information silos into an<br />
integrated information cloud, you can make<br />
the information much more usable, as it is<br />
easier to get to.”<br />
While NGA takes cloud computing seriously,<br />
the agency doesn’t have any unrealistic<br />
expectations about how easy it<br />
will be to deliver. “Placing GEOINT in an<br />
agile, high-performance computational<br />
cloud is certainly part of the strategy. However,<br />
that objective remains a vision and<br />
not yet committed to through the establishment<br />
of tangible acquisition programs,” said<br />
Cuppan. ✯<br />
Contact Editor Harrison Donnelly at<br />
harrisond@kmimediagroup.com. For more<br />
information related to this subject, search our<br />
archives at www.<strong>MGT</strong>-kmi.com.<br />
www.<strong>MGT</strong>-kmi.com <strong>MGT</strong> 7.1 | 13