Arm Community
Arm Community
  • Site
  • User
  • Site
  • Search
  • User
Arm Community blogs
Arm Community blogs
Servers and Cloud Computing blog A System Monitoring Control Framework (SMCF) for Arm Neoverse CSS
  • Blogs
  • Mentions
  • Sub-Groups
  • Tags
  • Jump...
  • Cancel
More blogs in Arm Community blogs
  • AI blog

  • Announcements

  • Architectures and Processors blog

  • Automotive blog

  • Embedded and Microcontrollers blog

  • Internet of Things (IoT) blog

  • Laptops and Desktops blog

  • Mobile, Graphics, and Gaming blog

  • Operating Systems blog

  • Servers and Cloud Computing blog

  • SoC Design and Simulation blog

  • Tools, Software and IDEs blog

Tags
  • infrastructure
  • Sensors
  • Neoverse
Actions
  • RSS
  • More
  • Cancel
Related blog posts
Related forum threads

A System Monitoring Control Framework (SMCF) for Arm Neoverse CSS

Marc Meunier
Marc Meunier
October 15, 2024
5 minute read time.

To achieve the highest performance while maintaining system stability, SoCs increasingly depend on monitors and sensors to deliver real-time actionable data. Specialized sensors can deliver data which can be analyzed locally for immediate action or provided for remote monitoring and analysis. In past systems, this was managed by a System Control Processor (SCP) with a view of the entire SoC monitors and sensors. 

Arm Sensor Framework specifies common sensors for runtime control processes across Client, infrastructure and Automotive SoCs. The framework includes Arm-defined as well as third-party digital sensors and monitors embedded into the cores. These sensors are used for system control functions for example minimizing overall system voltage droops. They are also used at the cluster level for power management and clock generation and for specific functions for example calibrating digital sensors.  

As systems scale with more cores, the complexity of managing sensor and monitoring data becomes an issue. Increasingly specialized sensors are needed which often deliver unique data to analyze and can come with different interfaces. Sampling data is becoming more flexible, adding to the need for more sophisticated control management. Third-party specialized sensors offer unique value propositions but have been traditionally difficult to integrate.  

System Monitoring Control Framework (SMCF) 

To address the increased complexity of managing sensor and monitor data, Arm introduced the System Monitor Control Framework (SMCF). The SMCF provides a standard software programming interface and comes with a set of functions to simplify setup. It enables 3rd party sensor data to be quickly integrated into a control system and allows developers to group monitors in functional blocks to create a manageable hierarchy.  

The SMCF adds several improvements and adds flexibility to the system:  

  • Enables sampling of multiple sensors. 
  • Support for internal and external triggers 
  • Sensor data dump into memory location for on-demand access
  • Configurable list of monitors 
  • Activity counters  
  • Capable of managing dynamically evolving system data
  • Flexible data width from 1 to 64 bits per monitor 

The SMCF consists of two components:  

  1. A Monitor Group Interface (MGI): a common configurable reusable component containing the software interface. The MGI connects to SoC-level interconnects and includes DMA capabilities to relieve the SCP from the overhead of pulling sensor data for analysis.  
  2. A Monitor Local Interface (MLI): an adaptation layer for specific sensor or monitor types. The MLI is where translations between specific monitors (sensor data) and MGI takes place.  

MLI blocks in a group monitor connect to one MGI as shown in the diagram below:  

Monitor group topology

Figure 1: Monitor group topology 

 

SMCF Interface Topology
Figure 2: SMCF for Monitor Diversity  

The SMCF supports different monitor management models. It also supports internally timed periodic sampling or interrupt driven by multiple external triggers. The selection of monitors to be sampled can be configured with a programmable delay for trigger-to-sampling.  

Sampling waveform
Figure 3: Sampling waveform

SMCF for Power Management

SMCF is a distributed framework that supports placement and management of monitors within SoC power domains. This removes the need for central coordination and indication of power domains states for monitor data collection. Hence, this allows power management with minimal, or no action required by configuring software. The MGI supports Sample IDs, written by the DMA interface, to indicate when monitors or even an entire MGI is powered off. It supports an interrupt event when an MGI is powered on to indicate that (re)configuration is required.  

Functionality that allows monitors to connect and disconnect from MGI. It also allows monitors to be powered off separately from their MGI without software support. The MGI/MLI power modes expected to be managed with AMBA LPI, Q/P-Channel along with other components in the relevant power domains.  
Power Management Central Collection (Non-SMCF Model)
 
Figure 4: Power Management Central Collection (Non-SMCF Model) 

 Power Management: Distributed (SMCF Model)

Figure 5: Power Management: Distributed (SMCF Model)

SMCF for Monitor Data Management

SMCF supports data collection directly from the group interface and through DMA to a memory-mapped location. Data availability can be supported through interrupts or through sample identifiers indicating data validity.  

For groups with more than one monitor, monitor sample data packing makes data collection more efficient. For example, a monitor that produces data values of less than 16-bits can have multiple values packed in a 32-bit location to make data reads more efficient.  

Sample identifiers for externally written data
Figure 6: Sample identifiers for externally written data 

SMCF Aggregated Data and System Telemetry 

The rich set of data collected by the SMCF can be utilized for system telemetry. This data can be accessed by the operating systems and used for profiling, debugging, benchmarks, and workloads. However, care should be taken to ensure that exposing the data does not introduce additional security vulnerabilities in the system. A standard interface to present suitably sanitized data gathered by the collection agent, for the consumption of operating systems and other agents is currently under design as an extension to the System Management and Control Interface (SCMI) specification. This helps with remote management use cases as well, where the collected SMCF based monitor/sensor data can be communicated to a BMC through a SBMR compliant interface to align with OCP guidelines.  

Integrations with 3rd third-party sensors 

With the layered approach of the SMCF, integrating 3rd third-party sensors has never been easier. ProteanTecs, a member of Arm Total Design, has integrated their monitors with the SMCF, providing an easy path for CSS customers to leverage their specialized sensors with minimal effort.

References   

  • System Monitoring Control Framework Architecture Specification: https://developer.arm.com/documentation/Den0108/latest/   
  • Arm Power Policy Unit Architecture Specification: https://developer.arm.com/documentation/den0051/latest/   
  • AMBA Low Power Interface Specification: https://developer.arm.com/documentation/ihi0068/latest/  
  • https://neoverse-reference-design.docs.arm.com/en/latest/features/power/smcf.html  
  • OCP Global Summit 2022 : Arm + Ampere presentation on “OCP HW Management”: https://drive.google.com/file/d/1Lf_W46LvXrDxtatMYv5a2NY99h-9_R-d/view  
Anonymous
Servers and Cloud Computing blog
  • Migrating our GenAI pipeline to AWS Graviton powered by Arm Neoverse: A 40% cost reduction story

    Hrudu Shibu
    Hrudu Shibu
    This blog post explains how Esankethik.com, an IT and AI solutions company, successfully migrated its internal GenAI pipeline to AWS Graviton Arm64.
    • August 28, 2025
  • Using GitHub Arm-hosted runners to install Arm Performance Libraries

    Waheed Brown
    Waheed Brown
    In this blog post, learn how Windows developers can set up and use Arm-hosted Windows runners in GitHub Action.
    • August 21, 2025
  • Distributed Generative AI Inference on Arm

    Waheed Brown
    Waheed Brown
    As generative AI becomes more efficient, large language models (LLMs) are likewise shrinking in size. This creates new opportunities to run LLMs on more efficient hardware, on cloud machines doing AI inference…
    • August 18, 2025