Is your IT infrastructure
 AI-ready? 

Once confined to the purview of science fiction stories, artificial intelligence (AI) has seen renewed interest in recent years with substantial progress in areas such as machine learning (ML) and deep learning. Today, AI delivers tangible business value when leveraged towards improving the customer experience, reducing cost and bolstering revenue generation.

According to analyst firm Gartner, the business value derived from AI is projected to reach US$1.2 trillion in 2018, a huge increase of 70 percent over the previous year. And as AI-centric solutions are increasingly utilised, this will impact not just existing hardware deployments but the design of IT infrastructure in the years ahead.

Ever wonder how will AI impact an organisation’s IT infrastructure? Here are the top 3 considerations in getting your organisation’s infrastructure AI-ready.

More powerful hardware

For a start, enterprises must look at more powerful computing hardware to process the huge volumes of data inherent to AI. While some experts think that current systems do not represent a cost-effective approach, existing hardware will have to do until the invention of specialised AI-centric solutions built on reconfigurable FPGA (field-programmable gate array) or ASIC (application-specific integrated circuit) microchips.

To establish and train various AI models, data must be stored and retrieved quickly at almost all stages of the process. This can only happen with high performance components such as GPUs (graphics processing units) and high-performance storage systems - the latter is essential for deep learning as larger datasets dramatically improves its accuracy.

With faster GPUs, larger datasets used, and the creation of more efficient model architectures, storage performance metrics such as IOPS (Input/Output Operations Per Second) is now more crucial than ever. One solution may be the use of solid-state drives (SSDs) deployed in advanced, all-flash storage arrays to replace current deployments of traditional rotating media.

Processing technologies should not be overlooked too, due to the increased need for compute capacity over time. Upgrading to newer microprocessors and fast SSD storage can help ensure that processing speed don’t get sluggish and overall performance remain consistently speedy.

Greater bandwidth demand

While the most rudimentary machine learning can be run on a single machine, more advanced AI processing is typically run across a network of computers consisting of scores of high-powered nodes. These machines are interlinked with high bandwidth connections to support continuous communication for AI training and data processing. Network latency matters too, as the scale of these networks means that poor latency can add up quickly to adversely affect overall performance of AI processing.

Thus, CIOs and network managers must consider how best to upgrade their existing networks with highly scalable bandwidth to support AI. Fortunately, a plethora of network offerings are available today, ranging from 10Gbps, 40Gbps, 100Gbps or even 400Gbps speeds that can be deployed as a digital highway to facilitate the algorithm computing between multiple servers.

Even for organisations that opt not to upgrade their network, an architecture refresh may still be beneficial for optimal AI processing. One of the common strategies is to compress traditional 3-tier networks that utilise multiple tiers of core, aggregate and edge switches to simpler 2-tier ‘spine-and-leaf’ networks. The latter allows servers to communicate directly and better supports the low-latency requirements of modern production systems and AI training networks.

Artificial Intelligence in the Data Centre

Despite the temptation to design an AI-centric network from scratch, the truth is that most processing within the data centre will remain run-of-the-mill applications and data processing with no AI. This makes it a wiser move to gradually move towards AI through a hybrid deployment, which has the added advantage of being able to support existing legacy equipment and systems.

Within the data centre though, AI can also be leveraged reduce an organisation’s carbon footprint through more efficient use of energy. For instance, AI algorithms can be used to autonomously tweak power and cooling systems with an eye towards greater energy efficiency. This daunting task is traditionally performed by data centre managers, who must monitor and manually adjust hundreds of thermostats across the entire facility.

On the other hand, an ML-based system can learn the optimal temperatures throughout the year to continually refine the algorithm for greater efficiency for significant cost savings. On that front, the Info-communications Media Development Authority (IMDA) in Singapore is working with National Supercomputing Centre (NSCC) on the use of AI to boost energy efficiency.

Elsewhere, AI-enabled robotics technology is another trend that is rearing its head. On the network, this revolves around automating the management of physical connections, reducing reaction time to security issues and increased productivity for IT professionals. IT managers are hence gain precious time savings that they can utilise to proactively manage the data centre and better support their organisations.

Is your organisation AI-ready?

While many companies understand AI's potential, most are still trying to understand how they might leverage it to meet their primary business goals.

For many of them, it may be more feasible to partner with external ICT service providers to tap into the benefits of AI without the need for extensive investments or lengthy implementation timeline. This could be in the form of AI-centric cloud platforms, or by establishing an infrastructure that integrates easily with turn-key solutions. With this approach, organisations can hence increase their business value through AI without having to take their eyes off their core competencies.

AI will undeniably be at the forefront of the next wave of business innovation. The wide-ranging benefits in terms of operational cost saving, the creation of new revenue streams, more efficient data-driven processes, and vastly improved customer experience are too appealing to be ignored. AI will bring more benefits to organisations due to the potential cost savings and your customers are likely to expect service providers to match up with seamless AI customer experience in the near future.

For more information on how StarHub can help to enable AI solutions in your organisation, please visit https://www.starhub.com/business/enterprise/solutions-and-services/digital-services/digital.html.

Follow StarHub Business on Linkedin for the latest business updates.