RZ Technology: AI is becoming a game changer

Artificial intelligence (AI) has arrived in the data center because its complexity is increasing so rapidly that it is often difficult to manage manually. “We are seeing significant growth in AI-driven data center automation,” said Paul Bevan, analyst at Bloor Research. “CIOs who ignore this will soon be left out,” Bevan continued.

AI is now being used for many purposes in data centers. Capacity planning algorithms exist to meet the demand for energy and required IT resources. Other areas are failure and anomaly detection, root cause analysis, and failure prediction. With IoT hardware and proper AI analysis of the collected data, many performance issues can be proactively addressed.

The use of algorithms in data center operations is not new. For example, automated backup devices, predefined configuration and deployment functions have been around for a long time. What separates new AI tools from these legacy automation facilities is the ability to “learn” and adapt based on many different factors, simplifying data in this way and dramatically increasing its reliability.

An important tool for increasing data center reliability is the detailed monitoring of all performance data of the devices in use. This is called telemetry devices Allows analysis of individual and custom components in near real time. The most important application area of ​​this telemetry is the detection of overloads on individual servers or components. This can happen due to device errors; But often there are extreme loads due to improper distribution of workloads. This then leads to resource conflicts, high processor temperatures, and similar problems. The larger the data center, the more frequent performance drops are likely to occur.

Reading hardware performance data allows you to quickly identify and isolate these issues and restart stopped processes. In addition, this telemetry data can also be used to fine-tune process management, because the specified measured values ​​can be correlated with the distribution of workloads over a longer period of time, so that predictive measures can then be implemented in order to avoid performance bottlenecks.

For this telemetry, the system components must have the appropriate facilities to generate the required measured values. For example, Intel® Xeon™ Scalable processors have corresponding performance monitors that monitor clock frequencies, cache usage, and similar parameters. The Intel telemetry aggregator (ITC), which provides the most important measured values ​​for power consumption, memory usage, or resource usage, enables you to start telemetry. However, for more efficient use of telemetry data, especially with larger server clusters, it is advisable to set up your own scalable software package that can collect, store, classify and evaluate data across all the clusters used. Then corresponding expansion using artificial intelligence enables forward-looking measures and automatic coordination.

Intel just got a road map Published for the use of artificial intelligence in the data center. An important pillar of this is the 4th generation Intel® Xeon® Scalable processors, which should be available from fall 2022. This is the most feature-rich Xeon processor with up to a 30x performance increase in all AI applications.

In addition, Intel already has a vision for the data center of the future (DCoF) advanced. This includes Intelligent Fabric, a new framework for Ethernet connectivity that aims to optimize all data centers from cloud to edge, and improve connectivity for cloud-native and remote applications. It combines a comprehensive suite of different connectivity, networking, and chip technologies within the framework of Intel Infrastructure Processing Units (Intel IPUs) to deliver intelligence, performance, scalability, and manageability.

You can learn all about Intel data center automation here