Harvesting grid computings power
The virtues of grid computing have been discussed before.
Now Satyam's grid computing practice offers consulting and other related services
for enterprises that want to set up an in-house grid computing infrastructure,
says Anil Patrick R
Your typical organisation has underutilised resources on its network. Grid
computing (See box A Grid Computing Primer to learn more about the basics of
grid computing) is a technology that can help enterprises optimally utilise
these resources. This is where Satyam Computer Service Ltds (Satyams)
grid computing service comes in; the company consults, deploys, manages, and
maintains enterprise grids for organisations that wish to have their own in-house
The grid computing practice is a part of Satyam's Consulting and Enterprise
Solutions unit. It offers consulting services to help organisations assess utilisation
levels of existing IT infrastructure. "The grid computing practice then
suggests measures to improve resource utilisation using this technology. It
also helps organisations plan for their future IT infrastructure and investments,"
said G B Prabhat, Director, Consulting and Enterprise Solutions, Satyam.
Grid computing's unique value proposition is the promise of delivering massive
computing capabilities using computing resources that are already present in
the enterprise. Computing tasks are distributed among existing resources on
the network and then reassembled after processing with this approach.
(Check Figure: Satyams Basic Grid Architecture to see the basic grid framework
used by Satyam.)
The distributed processing approach makes it possible to derive tremendous processing
potential from existing resources. This is why it is common to see grid computing
being adopted for processor-intensive applications such as genetic and space
research. Satyam is playing to the technology's strengths by targeting those
verticals that run compute-intensive applications.
According to Prabhat, the industrial verticals which will benefit from grid
technology include life sciences, geo sciences, financial services, government,
industrial manufacturing, and entertainment & media.
Tapping hidden strengths
processing architectures can provide considerable business benefits. Apart from
obtaining a better return on income from existing investments, grids can reduce
time-to-result, provide productivity improvements, and improve IT utilisation.
The technology can help avoid the pitfall of over-budgeting.
According to Satyam, one of the biggest benefits of using grid processing is
the marginal use of additional hardware. Organisations can build grids on their
own premises. As grids tap into the surplus processing power of existing computers,
the only additional hardware required for a basic grid implementation is a server
to host grid application software. Additional grid servers need to be deployed
for larger grids.
On the software front, grid application software is required to build a grid.
This has a server component residing on the grid server, and an agent component
residing on every node of the grid. "The additional requirement on hardware
and software to build a grid is very low and will be negligible when compared
to the cost of an High Performance Computing (HPC) set-up that equals the computing
power generated by a grid," said Prabhat.
Satyam demonstrates proof-of-value to interested organisations using its in-house
grid model at Chennai. Organisations can run a sample application on this grid.
The 200 node grid, across two locations, models a typical enterprise network.
Satyam maintains that by installing the grid computing hardware and software,
they have been able to simulate a live grid environment with 500 Gigaflops of
The company says that it is collaborating with leading product vendors to augment
its competency and extend its reach globally. It also has an alliance with United
Devices, a US-based grid solutions company, and is discussing partnerships with
other leaders in the field.
In addition to consulting, Satyam offers other-grid associated services. These
- Design of applications to run on the grid.
- Implementation: These services include building
the grid across available IT resources.
- Deploying the grid-enabled solutions to run on the
- Managing the grid for effective resource utilisation:
This will also involve making more applications run on the grid as well as
expanding it by bringing in more computing resources on the grid.
- Maintaining the grid: This includes maintenance
of the production environment to ensure availability and proper functioning
of the grid. It will also involve tracking the performance of the various
grid components, and meeting specific requests from the enterprise.
Satyam's Indian grids
A grid project has been executed for Dr Reddy's Laboratories (DRL). The objective
was to grid-enable the molecular docking application used at the pharmaceutical
DRL's biggest challenge with the molecular docking application was the restrictions
imposed by conventional systems. With the long duration that these systems used
to take, there was a restriction on the number of molecules that could be processed.
"Satyam deployed the application on the grid, developed interface programmes,
and submitted a database of 59,000 molecules to run on the grid of 50 nodes
built across two facilities that yielded phenomenal results," adds Prabhat.
The bioinformatics department of the University of Pune has also executed a
project in Satyam's grid lab.
Anil Patrick R can be reached at firstname.lastname@example.org
|Grid computing is a form of distributed computing that uses consolidated
utilisation of resources such as processing, applications and storage across
To describe it simply, a complex task, one that requires massive processing
power and storage capabilities, is divided into smaller tasks and sent across
to clients over the network (for example, LAN, MAN/WAN, Internet, etc) for
processing or storage; it is then recombined to get the final desired result.
This approach helps achieve massive capabilities at relatively economical
manners that are not otherwise possible with a traditional approach.
Looking at it closely, grid computing is not really a new
technology as such. One of the first implementations of this technology
was the UC Berkeley's SETI (Search for Extraterrestrial Intelligence)
project SETI@home. This project makes use of voluntary processing time
donated by users from across the globe to do the massive data crunching
required for deciphering extraterrestrial radio signals.
From an enterprise perspective, it has been a gradual evolution. The first
entrant was parallel processing with multiple processors (Symmetric Multiple
Processing, or SMP.) Then came massive parallel processing (clusters), and
after that came grids.
Then, as now, the typical grid architecture consists of grid server(s) hosting
the grid application and clients (nodes) which have software agents installed
on them. Enterprises have largely restricted themselves to employing the
grid over LANs. But this scenario is bound to soon expand to WANs and the
Internet, with grids becoming more popular in the enterprise. Service providers
who will build commercial grids as web services are also likely to materialise
along with this.