India must strive for ‘kotipakoti’ computing: Intel’s Raja Koduri

‘Kotipakoti’ is among the very big numbers that the ancient Indian mathematicians had visualised in the early texts. It is 21 ‘zeroes’ after ‘one’. In computer lingo, zettascale computing entails ‘kotipakoti’ computations that can give amazing insights by culling through gigantic volumes of data in real time.

Raja Koduri, Senior Vice-President, Chief Architect and General Manager (Architecture, Graphics and Software) at Intel Corporation, says it is quite achievable. “India can achieve it by 2025,” he said.

When it gets there, it will be a ‘Kotipakoti Ganananka Atmanirbhar Bharat’, achieving self-reliance in computing capacity.

“India can get there. Do not bother just catching up to the cloud, in which India fell behind. Just jump to the next thing. When India missed the PC drive, it jumped to mobile, where it leads,” he added.

Computational capacity

But why does India need such kind of massive computing capacities?

Koduri argued that India needs to cash in on the voluminous data storage potential in the country, and needs its data to be handled in local centres in order to benefit from it in real time.

He cited the example of challenges that farmers face. “There is so much uncertainty in terms of climatic conditions. There are issues related to insects, diseases, prediction of yields and marketing,” he said.

“Why is that the computation is only available to big companies? Why doesn’t the farmer have access to real time information? Why can’t they monitor and make adjustments to what they are doing on a real time basis?” he asked.

Huge computational capacities are also needed to get better outcomes in handling the Covid-19 pandemic.

“There’s not a facet of life where we don’t rely on computing. You need to have self-reliance in computing. It is as important as as self-reliance in food and electricity,” he told BusinessLine .

Koduri said it is quite important for the country to compute and analyse voluminous data that 130 crore Indians generate every day.

The country needs a zettascale computing capacity — 1,021 floating point operations and AI operations per second — to handle the data generated. “It seems pretty big but it is quite achievable,” he argued.

Computing capacity, like electricity, should be accessible to every one. “Some people paint a very dark future, saying computers will take over the world and it will be scary. Whenever a new technology is created, it is used for both bad and good purposes. But eventually, good purposes takes over bad purposes,” he says.

Better infrastructure

Koduri said there needs to be a collaborative approach to achieve massive computing capacities.

“We need to have multiple local datacentres where the data can be collected within a radius of 100 miles, so that you can collect the data in milliseconds. This will also help in addressing the issue of redundancy of data (same data stored, transferred across multiple devices). You don’t need to move the data across the country,” he said.

“It is like a national supercomputer highway. In five years, we are going to have supercomputers across all of India,” he said.

He feels the country should address issues such as electrical infrastructure. “You can’t have data centres be disrupted with power failure. We may have to build the generators and battery packs and other stuff,” he says.

Raja Koduri says Intel operates at all the levels that a basic data centre is configured to. “I call them the six pillars of a data centre. You need chips, architecture, memory, switches and networking product line, security and the software that runs all of this stuff,” he pointed out.

“Intel is one of the few companies where a country like India can just work with straight. We can be an exascale (1,018 floating point operations per second) or zettascale data centre partner. You can get all forms from transistor to software, from one place, designed securely,” he summed up.

COMMENT NOW