The integration of artificial intelligence into businesses is gaining momentum. As this integration accelerates, predicting the usage patterns of IT infrastructure resources for business applications is becoming increasingly challenging. While cloud computing allows for the scalability of IT infrastructure by enabling the usage of resources as needed, unexpected costs or security issues have led to deeper considerations by businesses when selecting IT infrastructure resources that align with the characteristics of their business applications.
Changes in IT Infrastructure Resource Usage Patterns for Business Applications
Traditional business applications such as ERP, SCM, and groupware have operated within predictable usage ranges. With the advent of the COVID-19 era, applications that heavily rely on network resources, such as online shopping platforms like Amazon, streaming services like Netflix, and video conferencing tools like Zoom, have surged. On the other hand, emerging AI models like ChatGPT, introduced last year, require significant computing resources.
The IT infrastructure resources needed for application operation consist of computing, storage, and networking. Depending on the type of workload, computing resources, storage, or network resources may be heavily utilized. Among these, computing resources consume a significant amount of electricity, leading to higher energy consumption. Increased electricity usage results in higher server heat generation, necessitating increased cooling capacity to maintain optimal temperatures in data centers. The usage patterns of IT infrastructure, focused on unexpected times or durations, are dynamically changing due to the concentration of resources for business applications.
Changes in Data Center Scale and Performance
Shifts in the usage patterns of IT resources for business applications and workloads bring about significant changes in corporate IT infrastructure investment strategies. Companies must now formulate strategies for IT infrastructure that can handle the rapidly increasing usage and growth rate of IT resources for fast-paced business applications, regardless of the scale or ownership and operation format of the IT infrastructure.
Scaling Up for Intensive Workloads
Data center sizes are continuously expanding. According to a 2020 report by the Uptime Institute, the global average rack power density increased steadily from 2.4kW per rack in 2011 to 5.6kW in 2017, and 8.4kW in 2020. This corresponds to a 3.5-fold increase and a 15% CAGR. Hardware with specialized processors like GPUs for machine learning can consume rack power ranging from 20kW to 50kW, which is up to 20 times the previous rack power. Companies are finding that rack power of less than 20kW is insufficient to keep up with the demands of modern workloads.
Ramping Up with Scalability
Demand for data centers with scalability is on the rise. Unlike the past when IT resource usage patterns were fixed, modern business applications dynamically utilize IT resources in terms of scale and timing. Data centers form the foundation and starting point in the value chain connecting data center infrastructure, IT infrastructure (computing, networking, storage), applications, and business. The scalability of data center infrastructure must satisfy two aspects: architectural structure and equipment layout. Architecturally, data center structures should allow for both horizontal and vertical expansion within the data center space to accommodate the physical placement of IT infrastructure equipment. Equipment-wise, mechanical and electrical systems should be expanded according to the phased addition of IT infrastructure equipment, enabling a cost-effective structure for initial investments.
Ensuring High Availability for Business Reliability
As the integration of business and IT strengthens, business competitiveness increases but so do risks. The primary role of data centers extends beyond supplying electricity and cooling to IT infrastructure in the server room; they must also ensure operational protocols to prevent supply disruptions. Even if a failure occurs in supplying electricity and cooling to the server room, backup systems should kick in through redundant supply lines to prevent interruptions in IT infrastructure operation. Redundant supply lines and operational protocols for backup should encompass not only electricity and cooling but also fire safety, security, and communication to ensure business continuity.
Connectivity for Various Network Options
Data centers must be designed to offer ample bandwidth and high transmission speeds, allowing for diverse network connectivity options. In addition to supporting graphics and multimedia applications like movies, games, virtual reality, and machine learning, data centers should offer network services tailored for secure operation of business applications where both speed and security are crucial.
Changes in Data Center Investment Costs
Increasing Investment and Operating Costs
The costs associated with constructing and operating data centers are rising significantly each year. This rise is driven by factors such as the performance levels (scalability, availability, security) of data centers, inflation, and, most significantly, the increase in electricity capacity due to rising rack power. As electricity capacity increases, the quantity and specifications of related electrical and cooling equipment rise, leading to higher capital expenditures (Capex). Factors contributing to the rise in Capex include geopolitical uncertainty, prolonged lead times in supply chains, inflation, and increased use of environmentally friendly materials for ESG compliance. Operating cost (Opex) factors include rising electricity and water fees, carbon taxes, and regulations related to major disasters.
The Evolving Role of Data Centers
Entering the era of the data economy, where people and things are interconnected online, data is accumulated, and AI is applied to analyze this data in a cyclic process, data centers will play a central role in facilitating data exchange, integration, and regeneration, thereby generating synergy in the data economy. While data centers are increasingly recognized as critical facilities, their importance is matched by the rising costs of investment and operation. Is it better to build and operate a data center directly or lease one? In the next post, we will explore the types and characteristics of data centers and discuss strategies for preparing data center infrastructure that aligns with our company's business direction.
Content is protected by copyright law and is owned by Daou Technology Inc.
It is prohibited to modify or commercially use this content without prior consent.
Featured images via gettyimages.
References
1. Rack density is rising, Uptime Institute, 2020
