Advertisment

High Performance Computing: High and Higher

author-image
DQC News Bureau
Updated On
New Update

With 25% growth witnessed every year and vendors opting for the channel route, partners have reason enough to smile.

Advertisment

Do you have computational tasks too large for a single machine? Are you

trying to model biological molecules, or render a 3D movie? Here is how computer

science can make your job easy. All thanks to one of its branches called high

performance computing or HPC.

To define it technically, HPC is a branch of computer science that

concentrates on developing supercomputers and also the software to run on them.

In layman's terms, HPC refers to a collection of computers that process a

variety of applications through both parallel processing as well as

multi-tasking in a very short period of time. HPC therefore implies a set of

machines working together to process an application and give results quickly.

Advertisment

HPC

can be divided into supercomputers and clusters. Supercomputers are employed for

specialized applications, which need large mathematical calculations. Examples

are weather forecasting, animated graphics, fluid dynamic calculations, nuclear

energy research, and petroleum exploration.

Clusters help to process huge amounts of data across networks within a few

seconds. Today, the HPC segment has moved beyond academia and R&D institutes

to include manufacturing, design, news and media, biotechnology, oil and gas,

pharma industries.

What's new?



It would be no exaggeration to say that the HPC concept is a relatively new

one and especially so in India. The latest in this area being, what is referred

to as Grid Computing.

Advertisment

In a broad sense, grid computing could be classified into three main types -

the campus grid, the metropolitan grid and the global grid. A campus grid would

essentially mean that two institutes are tied up in order to share both

resources and information.

Beuri 



Simmons

A metropolitan grid would, for example, handle data that pertain to the

weather forecasts of two regions- say Bangalore and Chennai, while the global

grid would have data exchange pertaining to weather forecasting around the

globe.

Advertisment

What criteria should solution providers consider while deploying HPC

solutions to their clients? It is important to remember that solutions provided

can be either CPU-technology dependent or dependent on memory-technology.

Also, with changing times, one is talking of enhanced memory in gigabytes.

Therefore, when one thinks of a solution, it is important that the solution

provided is scalable and more importantly available.

What that means, is that the solution must have the capability to match up to

the increasing needs of the user. Availability would mean that a scalable

technology is ready 'on demand'. These essentially are the key points to be

borne in mind while devising an HPC solution.

Advertisment

Challenges to overcome



Against the backdrop of an increasing need for HPC solutions, one needs to

keep in mind that demand will continue to ascend. Thus, one has to be prepared

to enforce betterment in the technology with time. The biggest challenges while

trying to do that are firstly the connectivity. This in essence means the need

for a dedicated bandwidth or pipe on the grid side especially, for smooth data

transfer.

Equally challenging are those issues pertaining to the processing end, which

could still do with a lot of betterment. Other challenges would be looking at

feasibility options of setting up the particular bandwidth. Another issue, at

least for the moment, is the huge costs incurred. Many HPC solutions run into a

few billion dollars.

HPC GYAN 

- DEFINITION:

Collection of computers that process a variety of applications through parallel processing as well as multi-tasking in a very short period of time



- RECENT DEVELOPMENTS: Grid computing can be classified into three main types - campus grid, metropolitan grid and global grid 


- OPPORTUNITIES FOR SOLUTION PROVIDERS: With vendors choosing to rope in channel partners, they can cash in on this chance for a quantum leap

Advertisment

Indian HPC is basically a vendor-dependant business and it is no exaggeration

to say that there is big money out there. IT vendors today, believe that HPC is

a high growth segment and can be adapted to any enterprise. The four major

players in this arena are Silicon Graphics Incorporated (SGI), HP, Sun

Microsystems and IBM. In India, companies like SGI and HP handle about half the

business in the HPC segment directly, while the rest is dealt by their tier-one

partners like: SIs, Tata Elxsi, CMC and others. Sun Microsystems banks on its

Indian partners Accel ICIM, Wipro, Locus, Summation, Skylark, Frontier and

Progression Infotech to deploy its HPC projects.

Business scenario



The market opportunity available in India is roughly about $30 million every

year with an annual growth of around 20-30%. According to Beurie, "India is

largely dependant on this business from outside. Since the country has huge

resources in terms of manufacturing, IT-services and workforce, we would

certainly stand to gain, as the demand for HPC solutions grows bigger."

Future prospects and likely beneficiaries from HPC solutions could include

R&D, education,



defense and other government projects in addition to a whole range of
manufacturing and other product industries.

With high growth potential of the HPC segment, vendors are trying to make

better inroads into the market via the channel route. They believe that niche

expertise is needed in this space. Depending on the kind of implementations,

turnkey projects would be available in areas like networking, storage networks

and related software expertise.

Vendors also believe that solution partners may have to shift from box

selling to technology selling to make a difference in the field. If partners can

capitalize on this opportunity, it could well mean a quantum leap for

business. 

Beuri Simmons is Head, Enterprise Systems (Servers and Storage) of NewWave

Computing.

Advertisment