The growing demand for artificial intelligence (AI)-based applications and data analytics is fueling the need for more—and larger—data centers. The expansion of cloud computing and the pandemic-induced work-from-home situation are also factors driving growth in the market.
The global cloud computing market is projected to reach $832.1 billion by 2025, in part due to the increased demand for Software-as-a-Service (SaaS)-based collaboration solutions and video streaming services that resulted from COVID-19 restrictions. That directly impacts the data center construction market, which is projected to increase from $207.2 billion in 2019 to $308.7 billion in 2027 at a compound annual growth rate of 6.4 percent.
AI in the Data Center Enables AI at the Edge
The very applications driving the growth in data centers are also being used to protect them. Integrators and cloud service providers (CSPs) are tapping Internet of Things technologies to make their data centers more reliable. AI and machine learning allow systems to analyze existing use, model alternative configurations, and find efficiencies that would be slow if not impossible to tackle manually.
AI in the data center can monitor physical issues such as temperature, power, and cooling, but it also can analyze embedded systems to improve efficiency and predict failures. For example, AI can monitor and balance server workloads, not just for scalability but for reliability. When an AI-enabled scheduler detects that a node is overheating, it automatically can shift the workload to another node before problems arise. The same is true for degrading memory.
Machine learning can also protect against security breaches. AI can learn normal network behavior and detect anomalies from cyber security threats or malware. AI-based cybersecurity systems also can identify potential loopholes in data center security and monitor incoming and outgoing data for possible breaches. The use of AI leads to increased uptime and better service level agreement protection.
Outages are costly. In 2020, nearly one in six data center outages cost more than $1 million, up from one in 10 in 2019. Memory failures are one of the three most common hardware failures that occur in data centers.
Intel (and) Insyde
Technologies such as Intel® Memory Failure Prediction (MFP) can protect against that. Intel MFP uses machine learning to compare error detection and correction data across multiple CSPs to develop a database of predictive patterns for system memory. It analyzes server memory failures at the micro level, and then uses that data to predict potential failures.
Intel MFP’s predictive capabilities give data center supervisors time to balance workloads, isolate defective memory cells, or replace faulty DIMMS before they trigger a catastrophic event. It is integrated in the Intel® Xeon® processor.
Intel® IoT Solutions Marketplace partner Insyde Software recently integrated the Intel MFP in its Supervyse Memory/FP firmware, a predictive memory failure solution designed for data centers and CSPs. The Supervyse Memory/FP uses the Intel database to assign a memory health score to each memory module and predict potential failures in real-time. It can be integrated with Insyde’s Supervyse BMC, InsydeH2O BIOS, and InsydeH2O UEFI firmware.
Our collective reliance on cloud computing and SaaS platforms will continue to climb. Service providers will require equipment with embedded IoT technologies to ensure their data centers offer high reliability and high availability.