Markdown Version | Session Recording

Session Date/Time: 15 Feb 2024 16:00

EIMPACT

Summary

The EIMPACT interim meeting focused on surveying existing drafts related to environmental impact in networking, identifying gaps in current work, and discussing potential future directions for the IETF community. Presentations covered various aspects from foundational challenges and metrics to specific Yang models and protocol extensions. A key theme was the balance between immediately actionable work and longer-term, more complex sustainability goals, along with the need for better communication and collaboration with other SDOs and policymakers. A strong sense emerged that while visibility and measurement are foundational, more work is needed on active control, configuration, and practical implementation, while also considering the real-world implications and trade-offs of sustainability efforts.

Key Discussion Points

• IETF Not-Well Reminder: Participants were reminded of the IETF Not-Well rules, particularly concerning IPR and conduct.
• Meeting Agenda: The agenda focused on reviewing existing drafts, understanding how they fit together, and identifying remaining work and gaps. The Gap Analysis slides were noted to be updated during the meeting based on discussions.
• Alex's Overview: EImpact-related Drafts and Landscape
  • Inventory of Drafts: An extensive list of EImpact-related drafts was presented, noting their scattering across various IETF Working Groups (e.g., NMRG, OpsAWG, IV) and the Independent Stream, alongside drafts yet to find a home.
  • Synopsis of Key Drafts:
    • draft-alex-nmrg-green-challenges: Analyzes challenges and opportunities in green networking management, structured across device, network, protocol, and architectural levels, emphasizing the need for data models and instrumentation.
    • draft-alex-green-metrics: Proposes metrics for environmental impact, covering equipment, flow, path, and network-wide perspectives, and the need to account for energy mix and attribute usage.
    • draft-ietf-eimpact-sustainability-considerations: Provides general guidelines for protocol and technology designers, discussing trade-offs and consequences of energy non-linearity.
    • draft-ietf-eimpact-sustainability-insights: Collects use cases for sustainability metrics/data models and suggests an operational support system framework for network sustainability data.
    • draft-marx-opsawg-power-energy-efficiency-yang: Information and Yang model to quantify power and energy efficiency, including benchmark data, real-time observations, and sensor management, augmenting an asset lifecycle model.
    • draft-ietf-iv-power-management-yang: A Yang model providing extensions for power information and control, tying into network inventory work, with common objects like power save capabilities.
    • draft-bonica-6man-path-energy-traffic-ratio-api: Defines an API for a metric relating throughput to end-to-end energy consumption for service attribution, without defining measurement methods.
    • draft-gandhi-eimpact-icmp-extensions-environmental-impact: Proposes ICMP extensions to request and return sustainability-related data parameters.
  • Categorization of Drafts: Drafts were broadly categorized into Foundational/General (e.g., challenges, insights, considerations), Instrumentation/Visibility (e.g., metrics, Yang models), and Programmability/Control (e.g., ICMP extensions, APIs).
  • Observations and Gaps: The primary focus is currently on energy/power consumption, with less attention to other sustainability aspects (e.g., embedded carbon, power sources). Gaps exist in active configuration/reduction mechanisms compared to instrumentation/visibility.
  • Related Efforts: Brief mention of past Eman work (RFCs on energy management) and ongoing Time Variant Routing (TBR) as potential areas for cross-pollination.
  • Discussion on Alex's Overview:
    • A need for considering the trade-off between the cost of data collection (e.g., flow-based metrics, AI optimization) and its benefits.
    • The proposed categorization sparked discussion, with a sense that much of the work presented leaned towards "visibility/observability," highlighting a potential gap in "control" or "influence" over network behavior.
    • Chairs agreed to curate and maintain a comprehensive list of all EImpact-related drafts, including those in other working groups, for better visibility.
    • The importance of addressing embedded energy and considering different time scales (short-term operational vs. long-term asset replacement) for sustainability impact was noted.
• Marisol's Presentation: Latest Developments on Sustainability Drafts
  • Power & Energy Efficiency Draft Update: The draft is being updated to adopt a controller-centric perspective, focusing on structured attributes from internal proof-of-concept and MVP implementations (e.g., Cisco products), and seeking multivendor feedback.
  • Background Observations:
    • Companies are setting Net Zero goals for 2030, requiring a significant acceleration in emission reductions.
    • ESG reporting is increasing, with sustainability expected to be a top decision factor by 2026.
    • Regulations (e.g., PUE for data centers in Germany) are emerging, pushing for tighter efficiency metrics.
    • The IETF needs to align with financial, legal, and policy makers (ISO, IEC, ITU, ETSI, CEN/CENELEC) to influence technology.
• Yan's Vision: Looking Back from 2032
  • Yan presented a forward-looking vision from the perspective of 2032, highlighting achievements since the EImpact program started.
  • Background: Anticipated increases in energy price volatility, public awareness of idle machine power consumption, and ICT's crucial role in solving emissions across industries.
  • IETF Achievements:
    • Improved measurement precision (from 30% to 1%).
    • Machine-readable metadata for metrics, enabling transparency and comparison.
    • Standard RFCs for collecting and aggregating data across vendors and systems.
    • Attribution of emission data, integrating with economic systems and GHG protocol.
    • Protocols improved to tolerate low-power modes and dynamic network conditions, with advanced traffic engineering and orchestration.
    • IETF engaged with EU policymakers, aligning technical work with policy objectives, and raising public awareness about ICT's "handprint" (positive impact).
  • Discussion on Yan's Vision:
    • The need for IETF to engage with other SDOs (e.g., 3GPP, IEEE) who are developing hardware on which IETF protocols run, to integrate hardware advancements.
    • The role of ISOC in promoting sustainability within its agenda and the potential for IAB influence.
    • IETF hackathons were suggested as a practical venue for open-source development and execution of sustainability ideas.
    • Concerns were raised about competitive advantage vs. standardization of data, but the idea of standardizing common subsets was seen as a way forward.
    • The importance of interacting with policymakers was stressed, acknowledging that IETF's impact might be more through providing working components ("bottom-up") than dictating policy ("top-down").
    • A proposal for the IAB to produce vision RFCs related to sustainability was welcomed.
    • Engagement with ISO and operator communities (like RIPE) was also suggested.
• Ron & Tony's Presentation: Yang Model for Power Management
  • Draft Goals: To enable power management, report per-component power consumption, allow components to enter power-save modes (down, standby, up), model dependencies between components, and provide the network element with expected traffic information.
  • Key Attributes: power-used, power-save-mode (initially Boolean, with discussion of more states), power-save-capability, automatic-power-management, functional-dependencies, expected-input-bandwidth, expected-output-bandwidth.
  • Discussion:
    • The simplicity of the power-save-mode (Boolean) was questioned, with suggestions for more granular power levels, noting the complexity of modeling heterogeneous hardware capabilities.
    • The recursive definition of components allows for detailed modeling down to individual transistors.
    • Concerns about the reliability of remote power management, with suggestions for "double-commit" mechanisms (automatic rollback if not confirmed) to prevent network outages.
    • The relationship between binary power save modes and proportional power management capabilities in hardware was discussed, emphasizing the need to avoid simplistic models that could hinder actual sustainability gains.
    • Maintaining dynamic dependencies was identified as a potential overhead concern. The draft authors noted that dependencies are expected to be updated at lower frequencies.
• Carlos's Presentation: Sustainability Considerations for Inter-networking
  • Draft Goals: To serve as a baselining document for sustainability in internet technologies (terminology, design, trade-offs, implications, use cases) and recommend a "sustainability considerations" section for new IETF drafts/RFCs.
  • Discussion:
    • There was a strong sense that for the recommendation to be effective (e.g., forcing a new section in RFCs), the document might need a stronger publication path (e.g., BCP or Standard Track) than an IAB Informational document.
    • The document's broad scope (foundational, guidance, specific considerations) led to a common suggestion to split it into multiple parts, targeting different audiences and purposes. For example, a standalone definitions document, high-level architectural guidance from an IAB perspective on trade-offs, and a BCP for specific protocol design considerations.
    • The IAB program chairs indicated a particular interest in distilling the high-level architectural guidance on trade-offs between sustainability and other network properties (e.g., availability, performance).
    • The document would benefit from incorporating "10 principles" for sustainability (as previously discussed by Yan Limat).
• Jay Nam's Presentation: ICMP Extensions for Environmental Impact
  • Objective: To gain visibility into environmental impact information (sustainability metrics) in an automated way, serving as a foundation for green routing decisions.
  • Approach: Uses RFC 4884 (ICMP Extensions) to append sustainability metrics to ICMP messages (e.g., Time Exceeded, Echo Reply), with an initial focus on a 32-bit power metric.
  • Discussion:
    • Suggestions were made to align the power metric definition with existing Yang models (e.g., draft-marx-opsawg-power-energy-efficiency-yang).
    • A broader discussion ensued on the suitability of ICMP as a transport protocol for this information, considering its ubiquity versus limitations for complex data or security concerns.
    • A key point of contention was whether this data provides "actionable intelligence" to other networks. Some argued that power consumption data is not directly actionable outside an administrative domain, doesn't change significantly with traffic, and could introduce security risks. Others countered that it could be useful within specific administrative domains for OAM purposes or for correlated path analysis.
    • The need for clear use cases to justify such signaling mechanisms was emphasized.
• Gap Analysis (Yari & Sures)
  • Metrics: Urgent need to finalize and publish RFCs for visibility and observability metrics, focusing initially on energy but not forgetting other aspects of sustainability (e.g., embodied carbon). Open source initiatives are critical.
  • Other Gaps: The ability to influence (not just observe) network behavior, effectively tackling the complexity of power-save modes, making informed choices about transport protocols (considering openness, data size, integration), and establishing trust in global data sharing.
  • Social/Process Gaps: Avoiding "perfect is the enemy of good" by focusing on actionable, simpler solutions. Enhancing bi-directional interaction with policymakers and other SDOs, clarifying IAB's role in publishing certain types of documents (e.g., architectural guidance vs. WG-specific technical specs).
  • Open Discussion:
    • Further discussion on operational states as metrics, including their semantics across vendors and the need for standardized definitions.
    • A strong preference for immediately actionable, short-term engineering goals was reiterated, such as collecting existing router data and establishing simple common metrics.
    • The role of existing research (e.g., energy-aware routing) in informing IETF's work was highlighted.
    • The importance of finding suitable "homes" for new work, especially for operational metrics that might lack a clear WG.
    • The need to address unsustainable protocols/technologies within IETF by raising awareness and documenting trade-offs (e.g., between efficiency and other properties like resiliency).
    • The challenge of open-sourcing operational data while protecting competitive advantage, suggesting anonymization and reciprocal data sharing models.
    • Metrics should aim to enable more granular company ESG (Scope 3) reporting and user-side carbon awareness.
    • Over-provisioning was suggested as a key metric to track, as it often leads to less sustainable network designs (e.g., relying on excess capacity instead of intelligent traffic engineering).
    • The economic aspects of sustainability and regulatory compliance, while not directly IETF's purview, significantly influence technical decisions, underscoring the need for IETF to provide tools for informed decision-making.
    • Prioritizing dynamic energy consumption (reducing energy when load is low) was seen as a key immediate goal, emphasizing the need for robust, reliable solutions to encourage operator adoption (e.g., automatic rollback of changes).

Decisions and Action Items

• Chairs:
  • Maintain a curated list of all EImpact-related drafts, including those in other IETF WGs, to improve visibility within the EImpact program. This may involve manually adding drafts or encouraging authors to include "EImpact" in their draft titles.
  • Investigate whether any identified critical EImpact work currently lacks a suitable home in an existing IETF Working Group.
• Community:
  • Participants are encouraged to engage in IETF Hackathons to implement and test EImpact-related drafts, fostering practical understanding and refinement. (Marisol noted her team's ongoing hackathon work).
  • The community is encouraged to contribute to the discussion on engaging with other SDOs (e.g., 3GPP, IEEE, ISO, ETSI) and policymakers, potentially with IAB facilitation.
• draft-ietf-eimpact-sustainability-considerations Authors:
  • Revisit the document structure with a strong preference for splitting it into distinct parts that address different audiences and purposes. This could include a standalone document for definitions, a high-level architectural guidance document (potentially for IAB stream, focusing on trade-offs), and specific design considerations (possibly a BCP).
  • The IAB program will prioritize distilling high-level architectural guidance on managing trade-offs between sustainability and other network properties (e.g., availability, performance, duty cycle).
• All Draft Authors:
  • Consider how their drafts contribute to immediate, actionable engineering outcomes, even if starting with simpler solutions.
  • For metric-related drafts, consider linking to existing Yang models and aligning definitions to promote consistency.
  • For drafts proposing new signaling mechanisms (e.g., ICMP extensions), clearly articulate the use cases, administrative boundaries, and actionable intelligence provided, addressing potential security and privacy concerns.
  • When defining power states or operational modes, aim for explicit definitions that indicate their impact on performance and operational behavior to avoid ambiguity across vendors.

Next Steps

• The IAB EIMPACT program will continue to oversee and track relevant work across IETF, ensuring that sustainability aspects are considered and advanced.
• Focus will be placed on supporting drafts that provide immediate, actionable engineering solutions for improving network sustainability, while also fostering discussions on longer-term strategic goals.
• Continued discussion on the EImpact mailing list regarding the proposed Gap Analysis, particularly identifying any missing areas of work and prioritizing efforts.
• Further engagement with other IETF groups, SDOs, and policymakers will be explored to align efforts and maximize impact.
• The community is encouraged to provide feedback on the updated Gap Analysis slides (available in the data tracker) and to actively participate in future hackathons.