The telecom industry journey towards sustainable, power-optimized, and energy-efficient networks is not a single step. It involves a phased approach starting from a cloud infrastructure foundation. Based on engagement with industry and customers, Dell Technologies recommends the power efficiency use cases that this section describes. These use cases can be categorized into two groups:
- Use cases that involve cloud-native features that are available within a cloud platform such as O-Cloud
- Use cases that require workloads and an orchestrator such as Service Management Orchestration (SMO)
Cloud platform use cases
- C-state and P-state optimization: C-states are the sleep levels of processor cores, and P-state refers to processor frequency. Modern operating systems are intelligent enough to transition cores to higher C-states and P-states when the core is underutilized. An intelligent scheduler can optimize energy consumption by consolidating workloads onto fewer cores when underutilized, allowing other cores to enter higher C-states and consuming less energy. This can be done dynamically based on workload demands to ensure optimized energy savings and performance. Implementing C-state and P-state optimization techniques in cloud-based systems can achieve significant energy savings. This approach will not affect the system's performance as the cores can be reused when needed.
- Energy-aware scheduler: Today, cloud deployments and workload placement are primarily driven by resource utilization. However, based on workload design, high utilization does not necessarily mean power efficiency. Using an energy-aware scheduler in platforms enables customers to co-relate and move workloads to take maximum advantage of both resources and power. This scheduler also delivers the best TCO and performance for customers.
- Predictive capacity planning: In a challenging economic time, the telecom industry is looking for ways to optimize network investments by building highly efficient platforms that can share capacity across multiple tenants. Building platforms that can collect, analyze, and report real-time data helps customers plan for future capacity needs.
- Infrastructure power optimizations: The benefit of AI/ML is not derived directly from the platform itself, but rather from the ability to build intelligence on top of the data available through the platform. Dell’s newly tailored solutions enable early use cases to measure and optimize the power usage of platforms based on run-time data.
Workload and Orchestrator-driven use cases
- Cloud-native autoscaling: Autoscaling brings elasticity to network infrastructure so that customers do not need to over- or underprovision resources. These features ensure that resource utilization matches demand and that idle cores can enter various power-saving modes. Also, the resources are instantly available when there is a need for scaling out. This native cloud feature is a reliable way to optimize energy consumption without impacting performance or stability.
- Intelligent workload consolidation: This is an intelligent system that considers infrastructure utilization, application utilization, Service Level Agreements (SLAs), and the criticality of the application before initiating power-saving decisions like workload consolidation. Energy savings can be achieved by completely powering off the system or entering different power-saving states of idle resources due to workload consolidation. This approach can reduce energy without compromising network KPIs, SLAs, or the user experience.
- Intelligent workload migration: Traditionally, hardware profiles for telecom workloads are static and set for maximum performance, even during periods of low utilization. However, it is possible to design clusters with multiple hardware profiles and use an intelligent system to migrate workloads to hardware resource pools during periods of low utilization for maximum energy savings. This approach can reduce energy consumption and costs while maintaining high-performance levels. An intelligent system powered by AI/ML algorithms can predict periods of low utilization and proactively migrate workloads to energy-efficient hardware, which ensures efficient hardware resource usage and reduced energy consumption.
- Energy-aware intelligent workload placement: Management systems like SMO can be equipped with an intelligent placement or homing system to place workloads on data centers and clusters with low carbon intensity. These systems can gather information about the nature of the energy source, such as fossil fuels or wind or solar energy, to understand the carbon intensity of the data centers and clusters and place workloads accordingly. Workloads that are not sensitive to latency or have geo-restrictions can be placed on these low-carbon data centers to reduce demand for carbon-intensive electricity. This approach can help achieve significant energy savings while reducing the carbon footprint of cloud-based systems.