data science

Safe home for the rising piles of data

Rising digitisation has made data centres the cornerstone of ICT landscape. We speak to industry leaders about tech options, challenges, policies, and market prospects

This is the age of digitalisation and the pandemic has only speeded up the trend. More people are turning to ICT for all that can be accomplished virtually: work, shopping, financial transaction, or entertainment. But every keystroke creates data, and increasing digitalisation means an increase in the data, each piece of which is potentially crucial for business.

The pandemic-pushed digitalisation thus translates into the rising demand for warehousing and maintaining a humongous quantity of data. That’s where data centers come into the picture: the dedicated physical spaces, usually buildings or whole parks, full of routers, switches, firewalls, storage systems, servers, and application delivery controllers. Their critical significance at the crossroads of business and technology cannot be overstated, and in the days to come India is going to need more and more of them.

We see a clear surge in hyper-scale data centres that have resulted in higher demand for resources such as power, water, skilled manpower and land.

— Sudipta Sanyal, CTO, Data Center Business, Sterling & Wilson

India’s data consumption will be about 25 GB a month per user by 2025, according to Ericsson Mobility Report, June 2020. That is a lot of data, and it will continue to grow, by machines and devices, rather than humans. Where this data is stored and analysed will be the key. Data centers (DC) will be at the epicenter of this disruption.

As digitisation supports adapting to the new normal and the emerging technologies are being utilised, an increasing volume of data is being generated. Data tonnage has been creating obstacles in the concentration and distribution of data. Edge computing is one of the solutions to keep a check on it.

A Gartner study suggests that by 2022, 70% of data will be created outside of the DC or cloud. With technologies like AI, ML, robotics, and automation becoming more common, data will be generated by machine-to-machine (M2M) technologies and will require large investments for it to be managed and moved. Therefore, data tonnage is definitely expected to be a real challenge, especially in terms of distribution and concentration of data.

DCS is expected to be the next real growth opportunity across the country due to the global pandemic, says Sudipta Sanyal, CTO, Data Center Business, Sterling, and Wilson Pvt Ltd. The lockdown has forced the nation to accelerate the shift to online, driving a massive spike in data needs.

The pandemic challenge is creating the need for data centers to provide rapid, high availability of data center services. This will enhance the speed of cloud adoption. More and more enterprise will encourage their employees to WFH. This will require better bandwidth connectivity for reduced latency. This will also enhance the application of edge to bring the services close to the demand.

The opportunities created by the pandemic are, however, also accompanied by new challenges. As with the whole of the technology sector, the data centre infrastructure industry is undergoing a transformation. The business scenario remains buoyant, but international investors may delay their plans, which can create hiccups in the short term. In the mid- and long-term, the investment would be back on track.

DCs are very high on the technology spectrum in terms of design, engineering, and execution. The need for a specific focus with high-performance teams that have relevant know-how on standards, equipment design, understanding of uptime and electrical and IBMS makes this space very complex. The pure DC EPCs have an upper hand on the generic EPC players as the industry clients are very demanding and involved in the complete value chain.

Sanyal says: “We can foresee fast-tracking of projects as the economy crawls back to normalcy. We are trying to mitigate challenges in execution by creating awareness among our site staff, providing PPE, and providing a conducive and safe environment. This will increase the service cost, specifically for the site works, which comprises a significant cost for the conventional brick-and-mortar data center. It will lead to more demand for pre-fabricated system and subsystem of DC infrastructures, including, modular pre-fabricated DCs.”

Data tonnage challenge gets real

Now that the pandemic has been here for nearly a year, do the DC players see the data tonnage challenge getting real?

Vinil Menon, Senior VP, Enterprise Applications, CitiusTech, says that in the healthcare space, there has been a profound shift towards telehealth, remote patient monitoring, and consumer-generated health data. The pandemic has accelerated the adoption of these technologies. Healthcare organisations globally are using cloud infrastructure, virtualisation, and other approaches to ingest, store and process massive volumes of patient-generated data coming through millions of connected devices. Data tonnage is a real challenge in the healthcare industry and will continue to be so in the years to come.

Sanjay Motwani, Business Head, Legrand Data Center Solutions (LDCS), agrees that data tonnage is now real. COVID-19 has come as a catalyst and led to a rise in data consumption in other online areas, such as entertainment, shopping, and education. Managing this and meeting customer expectations for quality experience is one of the biggest challenges.

COVID-19 has come as a catalyst and led to a rise in data consumption in the other online areas, such as entertainment, shopping and education.

— Sanjay Motwani, Business Head, Legrand Data Center Solutions

Nikhil Rathi, CEO, and Founder of Web Werks India adds that the question should more appropriately be phrased as: Are there enough data centers to handle the challenges of increased digital workflows? India remains a severely under-served market when it comes to data centers. We have a large, young, mobile-first population that uses technology on a daily basis. When you couple that with the pandemic situation, it requires greater use of technology. You find that we simply don’t have enough digital infrastructure to serve their needs at a level that is optimal.

“Yes, the challenges are real and we will have to speed up the availability of data centre space across the country. We are optimistic that supply will catch up with demand eventually, given that both the central and state governments are taking steps forward to incentivise the setting up of digital infrastructure across the country,” he says.

Santosh Kulkarni, Country Manager, Business Development, Smart Solutions, Vertiv India, says that while COVID-19 has played a significant role in the increased volumes of data being created, this trend had started before the pandemic. In early 2019, the World Economic Forum had suggested that the amount of data being generated was increasing exponentially and will touch 44 zettabytes by 2020.

AI race altering compute the equation

Going forward, it is quite likely that the AI race will alter the computing equation. Motwani identifies three factors driving AI: algorithmic innovation, data, and the amount of computing available for training.

The amount of compute used in the largest AI training runs has been increasing exponentially with a three-four-month doubling time (by comparison, Moore’s Law had a two-year doubling period). The trend represents an increase by roughly a factor of 10 each year. It’s been partly driven by custom hardware that allows more operations to be performed per second for a given price (GPUs and TPUs), but it’s been primarily propelled by researchers repeatedly finding ways to use more chips in parallel and being willing to pay the economic cost of doing so.

On the other hand, the cost will eventually limit the parallelism side of the trend and physics will limit the chip efficiency side. Given he precedent for exponential trends in computing, work on ML-specific hardware and the economic incentives at play, it would be a mistake to be confident this trend won’t continue in the short term.

Kulkarni says companies across the world have been adopting AI in various aspects of their businesses to ease the workload on their employees, and achieve faster results more efficiently. This also helps organisations analyse data and derive relevant insights. OEMs providing AI hardware to support businesses need to focus on creating solutions that efficiently compute data.

With AI, data can be handled, extracted, or analysed, for precise and quick results, notes Manish Israni, Executive VP and CIO at Yotta Infrastructure. AI makes products more efficient, hence enterprises must include AI-based innovations in their IT strategies. Various sectors have made use of AI algorithms to manage their data and get valuable insights and business analytics.

For instance, consider monitoring the securities market and fraud detection in the banking and finance sector or analysis and simplification of data collected by sensors and devices in healthcare. However, to boost AI/ML workloads, high-performance computing (HPC) also plays an important role. With HPC and AI/ML, the go-to-market for companies improves drastically.

Sanyal adds that while COVID-19 has been a ‘black swan’ event, it has been a defining moment for the cloud. Besides accelerating the pace of digital transformation, the crisis has also redefined people’s habits, compelling them to dive into the rapidly-expanding ocean of digital services. This, along with the increased use of AI, automation, robotics and predictive analytics by organisations, has led to an unprecedented demand for DC that supports cloud storage, and the wide spectrum of digital services run by governments and enterprises. Collectively, these play a pivotal role in keeping the lights on for businesses and supporting people to keep up with the ‘new normal’.

India remains a severely under-served market when it comes to data centres. We simply don’t have enough digital infrastructures to serve the needs at an optimal level.

— Nikhil Rathi, founder & CEO, Web Werks India

“Technically, more application of AI will demand a fast CPU with more cores and a larger implementation of GPU, which will increase RACK kW, demanding the planning of the electrical and cooling infrastructures for high-density data centers within the same footprint more efficiently. This will provide benefits in terms of space utilization, and higher speed and less latency for the compute-intensive workload,” he says.

Edge comes into play

Today, the DC industry is split, not about whether edge computing will be useful, but whether it will be profitable. Menon says that particularly in the healthcare space, the IoT ecosystem is flourishing. This is partly sensor and biometric data that has been in use for some time now. We can easily understand and incorporate new use cases arising from patient-generated data and consumer apps.

Another aspect unique to the healthcare industry, especially in advanced markets like the US, is the development of value-based care models that depend on real-time streaming data to be effective. Within these markets, there is likely to be sufficient demand for edge computing to make data sharing and data processing faster, more efficient, with the added benefits of security and privacy.

Kulkarni says companies across the world are increasingly seeing the benefits of adopting cloud-based technologies and edge computing. Edge computing has positively impacted the productivity and growth curve of companies, especially in times of the new remote working. It has enhanced security and helped reduce costs for businesses.

As companies increasingly move their workload to cloud-based technologies, the overall edge computing market is expected to reach USD 13 billion globally, according to a Gartner study. Another report by Grand View Research states that the market volume of edge computing is expected to touch nearly USD 43.4 billion by 2027 globally, with a crucial role played by technologies like 5G and IoT.

It is safe to say that while edge computing has proven to be useful and is gaining popularity, there is no doubt as far as the profitability of the market is concerned. With the surge in companies adopting cloud-based solutions, the market is bound to increase in volume and size.

However, Rathi notes that edge computing in India is still some way off. At the moment, it is simply not profitable, as services that rely heavily on edge computing haven’t been implemented in most parts of India. These services will take place eventually and customers will also require edge computing at that point. The economics will make it a necessity. DC players will have no choice but to move with the times and deploy edge DCs.

Motwani adds that whoever looks at edge computing shouldn’t look at each site as an independent site. It has to be visualised as part of a larger network, and what is the mileage one is getting by extending one’s reach using an edge DC. It is a business decision. If you are not to expand to an extended location with an edge DC, you may impact your overall business. Hence, the edge DC has to be visualised as a part of a portfolio. Moreover, if one is able to offer the right value, one can monetise the opportunity well.

Customer demands, market competition and the perpetual change in the dynamics of the DC industry are keeping the players on toes. Innovations are needed.

— Manish Israni, Executive VP & CIO, Yotta Infrastructure

Sanyal notes: “We see a clear surge in hyper-scale DCs that has resulted in higher demand for resources such as power, water, skilled manpower, and land. The scarcity of these resources has led to a shift towards the cloud for IT services. Many enterprises are shifting to clouds for IT services, as they like to opt for the Opex model, rather than the Capex model. This trend will increase multi-fold post-COVID-19 as more and more employees adopt WFH. This is helping to de-risk the business.”

The other benefits that CIOs are getting are out of regular day-to-day IT operations and maintenance, and devoting quality time aligning with business needs. This trend is increasing the demand for hyper-scalers like Amazon and Microsoft, who are setting up large DC facilities across various geographies. These hyper-scalers also setting up DC in colocation facilities increasing demand for colocation providers.

Sanyal adds that emerging technologies such as Big Data, AI, ML, and IoT will be the key drivers in increasing demand for the data storage system and hyper-converged infrastructures. Eyeing the growth potential, the government might look at establishing specific DC parks with prescribed power and water availability and tax incentives to attract foreign investment and accelerate growth. This may help in setting up campuses in class B and C cities.

The 5G rollout will accelerate the demand for edge computing. Automotive and large-scale manufacturing industries are expected to take the position of early adopters. Application of remote, secured modular and micro DC is among the emerging trends for the data center infrastructure.

“Edge will be an essential part of this hyper-scale environment for last-mile integration to meet the needs of the industry for distributed computing, storage, and better connectivity for reduced latency and increased client delights. Hence, the success and profitability of edge cannot be separated from the complete hyper-scale environment,” says Sanyal. That leads to hyper-scaling. We need to look at how the hyper-scale companies are deploying their own fiber and subsea cables, and SDN-powered cloud connectivity.

Motwani says that hyper-scalers have had their foot in the fiber and subsea cables for a while. With the amount of traffic that they are now carrying and its potential in the future, it makes more sense for them to have their own infrastructure, and optimise it as per the type, volume, and nature of data that they carry. They do not have to share all this with third parties, thereby maintaining greater security and confidentiality.

Carrying data through a network requires some hardware and some software. Splitting the software from hardware and further splitting the software into the control plane and data plane, and storing into the cloud makes the most logical sense. By doing this, hyper-scalers are able to manage or tweak the element which needs focus. At a basic level, it allows for greater responsiveness and flexibility. The result is cost-effective agility for customers.

There is also a need to manage this software-defined everything in the data tsunami. Menon says that many of their healthcare customers prefer to migrate to best-in-class cloud vendors to drive their SaaS and micro-service strategies.

Data centre modernisation

Robust DC modernisation plans are taking place now. According to Israni, customer demands, market competition, and the perpetual change in the dynamics of the DC industry are keeping the players on their toes. Various IT and conventional businesses have declared permanent remote working for a proportion of their workforce. Innovations are needed to enhance the performance, meet their increased requirements, and keep up with the modern upgrades to retain the customers, as building the right IT infrastructure is among the top priorities of the enterprises.

The micro data centre market size was valued at USD 4 billion in 2019 and is expected to cross USD 15 billion by 2026. There are immense growth opportunities.

— Santosh Kulkarni, Country Manager, Business Development, Smart Solutions, Vertiv India

At the same time, enterprises are now jumping to the cloud rapidly. This calls for the latest and updated hardware infrastructure to serve their needs. What differentiates a quality infrastructure from the regular one are the certifications and tier rating. An Uptime Institute Tier IV-rated DC will score over others in terms of robustness and providing failsafe infrastructure.

“Apart from IT systems, other DC infrastructure, such as cooling systems and power sources need to be robust and redundant. For instance, as the youngest and latest entrant, our DC features the latest and updated hardware/servers. We also use Li-on battery banks for UPS, and these are more power-efficient and have low recharge cycles. As a DC provider, improvements, or enhancements to make the IT facilities easier to maintain will be ongoing. It offers a better-managed infrastructure to our customers and brings in cost efficiency for us to operate,” Israni points out.

Sanyal says that virtualisation, the most important aspect of DC modernisation, has enabled the creation and deployment of applications faster and store data easier than ever, which has helped grow data exponentially. The more the data gets generated, the more important it is to preserve and protect it with backup and replication, driving the demand for storage. These results in serious challenges for IT departments, especially for those who want to consolidate IT infrastructure with cloud-based applications, virtualisation, and file sharing. In such a scenario, rebalancing is of utmost importance.

He adds that DC companies are achieving the rebalancing by capacity augmentation, within the same space, and performance enhancement. This means that data processing time can be reduced from hours to just minutes. Thus, organisations need fewer servers, hard disk drives, and fewer software licenses.

There is compatibility to create a truly unified storage environment. Usability (fit for purpose) gives you the ease of use, to optimise virtual machines with just a few clicks so you can easily deploy as many shares per minute, not hours. There is reliability. Look for systems that have no single point of failure architecture and data protection. Look to keep your applications online to improve recovery in the event of a problem. All these need to pair with a meticulous study on the cost economics and financial viability of the strategies in consideration of the problem at hand to get a winning solution.

On the physical infrastructure front, a lot of developments are taking place in form of the use of clean fuel and renewable energy, application of liquid-cooled servers, more granular infrastructures for making the Industry more sustainable and viable both environmentally and commercially.

The market for micro data centers

As the history of the size of mobile phones shows the trend is always towards making things smaller, faster, better, and cheaper. This trend will continue in the DC market with micro data centers. Having said that, India has not reached that stage, just like the case of the edge DC. Rathi says that the market for it at the moment seems limited to university or government research centers that require high-performance computing, the military, and perhaps, certain organisations that require it for R&D purposes.

Kulkarni says that a micro DC is a flexible and quickly deployable solution. Moreover, they are within the vicinity of commercial spaces, are easily portable, and can be managed remotely. Micro DCs are available in the ranges of a single rack DC of 3.5kW to a range of multiple rack solutions of up to 90 kW. They are mostly customised to user requirements and computing needs along with modular expansion and upgrades.

“The compact and portable nature of micro DCs helps organisations carry out efficient management and data security. The micro DC market size was valued at USD 4 billion in 2019 and is expected to cross USD 15 billion by 2026. There are immense growth opportunities, especially with the kind of technological advancements we are witnessing,” he adds.

On the other hand, Menon notes that the scale and cost advantages that hyper-scale DCs offer is tougher to achieve using micro DCs. There is an increasing focus on privacy and security, especially, with more organisations wanting to do more with their data as part of their ML initiative. When combined with privacy concerns, this space will benefit tremendously from micro DCs. Often, the comparison will be made between smaller DCs and simpler to manage private cloud or hybrid cloud variants, when it comes to meeting custom storage, network, and computing needs.

However, a modular solution, for example, hyper-converged infrastructure (HCI), makes it easy to replicate and deploy, especially if you are an ISV in a regulated space, and where compliance is a key requirement. Within these markets, there is likely to be sufficient demand for edge computing on public/hybrid cloud to make data sharing and data processing faster, more efficient, with the added benefits of security and privacy.

Motwani believes that while vendors have made products available for the micro DC, providers will have to educate the consumers about its benefits and advantages before we see real growth in this segment. MDC is more for the business side, rather than the IT side. Its application has to be thoroughly understood before we see real growth in this space.

Sanyal also feels that micro DCs are a growing market. The global market is relatively small in terms of value, but growing at an impressive rate of 52.7% as per a study by 451 Research. Post-COVID-19 growth will be much higher.

Compactor micro DCs are mostly application-oriented, and hence edge or micro DC as a proposition must be very strategic. Any company needs to first define its operation methodology, and then make a go-to-market combination of conventional and edge or micro DCs, which is optimal to meet its business goal. An edge or micro DC can cater to applications where latency needs mitigation. At the same time, it needs to exchange data with a conventional setup to upload the generated data.

What next, data center districts?

A whole district of the data center is a concept that’s time has come. In the union budget 2020, the government outlined the plans to help enable private sectors to build data center parks in India, a market that is likely to reach USD 4 billion by 2024, according to government estimates. The new policy is expected to incentivise setting up data centers, similar to some US states that have relaxed taxation on DC providers. In the past few months, there have been many companies that have opened their data centers in India.

Reduction in latency is also becoming a criterion for better user experience when users are accessing various online platforms. This will trigger distributed infrastructure development closer to users. Thus, new DC districts will get developed based on user demographics and data consumption patterns.

Motwani says that different states are at different levels of realisation of their DC plans. Due to COVID-19, a lot of state funds have to be diverted to healthcare, and that may have impacted their DC district plans a bit. The changing economic situation has also shifted priorities for different states. There is a clear shift happening, the timeline of which cannot be predicted at this point, he adds.

Sanyal says that for region planning a DC district should have enough power availability, no water scarcity, access to talent, and good connectivity. Looking at these filter points, there are not more than four or five regions in the country, like Navi Mumbai, which has close to 40% of the commissioned DC capacity and is growing at a rapid pace, followed by Chennai that has an abundance of land at affordable prices, besides proximity to SE Asia.

The other likely candidates are Hyderabad, Bangalore, and Delhi NCR. Although these regions have most of the required elements, a lot more needs to be done on grid stability, fiber-optic connectivity, and incentives to attract further investments.

The state governments need to play a critical role in providing the infrastructure for the DC industry. Since sustainability for this industry relates to its ability to move fast, it is essential for governments to facilitate in creating these infrastructures and provide real single-window clearance based on the success stories in many countries in SE Asia and China.

Healthcare organisations are using cloud infrastructure, virtualisation and other approaches to ingest, store and process massive volumes of patient-generated data.

— Vinil Menon, Senior VP, Enterprise Applications, CitiusTech

DC districts are already forming up as we speak Rathi points out. Though not in the shape of a formal DC park created by a state, there are clusters DC players have started building. A stretch in Airoli, Navi Mumbai is home to a number of DCs, and more are coming up. This is thanks to several factors, such as land availability, power availability, connectivity, and locations that are less prone to disasters. According to Israni with the increasing investment in DC by real estate players and the awareness of its economic benefits, the government has been supportive and focused on creating the digital infrastructure for the future.

Remote, unstaffed DCs next?

Again, the crisis that has created the opportunity is not without challenges. Given the mobility restrictions that led to WFH becoming the norm in other industries, will there be unstaffed DCs? Rathi says that remote management takes place in some form already. However, the day of unstaffed DCs is still a long way off. DCS is simply too critical to be left without human oversight. Some manual tasks have been automated, but in order for a DC to be totally unstaffed, it would require a level of automation, AI, and robotics that today seems more like science fiction.

Adds Israni: “DCs are meant to function without the interference of humans, once of course, the IT infrastructure is setup. COVID-19 has highlighted and made remote management of DCs necessary. Companies that had their own captive DCs, that is, on-premise, were most affected as they could not access their set-up during the lockdown. Moving to a multi-tenant third-party DC ensures a minimum service level for the hosted set-up.”

Companies have been working towards creating a DC environment that would work smoothly with minimal human intervention, as the number of servers to be taken care of is way more than the staff on the ground. Machines or automation will not impact the importance of the workforce at DCS. They add value and give you a more controlled environment.

Kulkarni says the social distancing norms have forced many companies to have minimal staff managing their DCs. Therefore, companies have had to adopt technologies like AI and ML and automate processes accordingly, to adjust to the new normal. The new DCS, or the ‘lights out’ DCs, are being managed by automation, making the process more efficient, reducing the need for manpower, and ensuring employees’ health safety.

“As DCs are being monitored remotely, ML plays a crucial role in their management. Technologies like AI and ML will provide the companies with all the information and insights about their DCs, right from the amount of power bring used to any anticipated risks. At Vertiv, we have strong solution offerings in the area of remote monitoring of DCs,” he says.

Motwani says the DC floor on its own has been pretty empty. It is always about the people in the building, working on various projects that are accessing DC. How much of the on-site staff would you move to remote working is the question. Depending on the kind of applications and processes the data center is running, there is a potential security risk element with remote access.

“Given the current times, there has been a need to shift some of the routine tasks of DC management to the remote environment. So, yes, we will do more remote management, and this will need special tools like KVM. Being a strongly secure product allows data center administrative staff to access the data center securely and remotely,” he concludes.

2 comments

  1. Ramkaran Prajapati

    Hello sir and madam i study in twelfth class i am interested in this course of ISRO my interest is more science geography and maths i also want to do this course. This is a golden opportunity for all of us. I also want to contribute to ISRO by doing this course.
    Please reply me.

    Thank you sir or Madam

Leave a Reply

Your email address will not be published. Required fields are marked *