SVTA_LOGO_FINAL_revised_2022

Edge Meeting Minutes
April 9th, 2026 9:00am to 10:00am PT</p

Attendees

Yoav Gressel (Streaming Video Technology Alliance (SME))
Alan Arolovitch (2you.io)
Piotr Galecki (Siden)
Ben Rosenblum (Vecima Networks)
Glenn Deen (Comcast/NBCUniversal)

Minutes

The meeting focused on discussing request routing and load balancing strategies for edge caching workloads, with particular emphasis on whether HSP (Hosted Service Provider) or tenants should control routing decisions. The group debated two approaches: one where tenants retain full control over routing and another where HSP acts as a middle-layer load balancer, potentially offering optimized delivery across edge networks. Glenn raised concerns about creating a “middle child” role between network operators and tenants, while Alan suggested keeping the solution simple and minimizing exposed controls since tenants might prefer to maintain control over their routing strategies. The discussion highlighted the tension between control and trust in the proposed architecture, with considerations about whether to offer sophisticated load balancing features or keep the system simple. The conversation ended with plans to continue the discussion in future weekly calls and to prepare messaging materials for upcoming presentations, including a demo at NAB.

 

Action Items

  • Alan: Send a reminder to JT about the survey status and CC Yoav and Sanjay
  • Florencia: Monitor and follow up on Edge Foundation document voting, and notify the group once quorum/voting is complete
  • Alan: Try to put together messaging/content for the Edge Working Group 101 presentation, to be discussed/reviewed on the next call
  • Glenn: Share previous presentation materials/talks about the Edge Working Group vision for review by the group

Antitrust Statement and Group Discussion

Survey and Routing Discussion Update

  • The team discussed the status of a survey that was sent to JT but not yet distributed to the full membership.
  • They reviewed the request routing options, focusing on whether HSP or tenant should handle routing and associated failover logic.
  • The discussion covered the need for tenants to indicate their routing approach and workload types, particularly for handling different types of workloads in the same location.
  • Florencia provided an update on the Edge Foundation document ballot, noting they were short a few votes but expected to reach quorum by the following week.

Cache Load Balancing Strategies Discussion

  • Alan and Yoav discussed load balancing and failover strategies for cache instances across multiple sites.
  • Alan emphasized the need for sophisticated load balancing to prevent capacity overruns and cascading failures, suggesting the use of round-robin distribution and weighted load balancing based on user location preferences.
  • They agreed that both load balancing and failover should use the same data model but serve different purposes, with load balancing focusing on optimal capacity utilization when systems are healthy.

System Failover and Load Balancing

  • Alan and Yoav discussed failover and recovery strategies for managing system capacity and load balancing across multiple sites.
  • They explored different approaches, including site redundancy and load balancing between adjacent sites, with Alan emphasizing that not all solutions need to rely on internal site redundancy.
  • Yoav suggested defining failover cases based on insufficient capacity to handle current loads, while Alan noted the existence of established mental models and solutions in global load balancing that could be adapted.
  • The discussion concluded with Yoav asking for thoughts on the minimum complexity required for their approach.

HSP Request Routing Implementation Discussion

  • The team discussed implementing caching workloads in their infrastructure, with Yoav explaining that while tenants could handle their own routing, there was value in providing HSP (Hosted Service Provider) request routing as a service. Glenn raised concerns about potentially creating parallel but misaligned behaviors that could conflict with different hosting workloads.
  • Yoav clarified that while caching workloads typically come with their own strategies based on central deployments, the HSP approach could provide better solutions and strengthen partnerships with network operators by leveraging their deeper knowledge of the networking topology.

Edge-Hosted Environment Models Discussion

  • Glenn and Yoav discussed different models for edge-hosted environments, with Glenn proposing a smart overlay model that would handle optimization automatically without requiring users to manage specific details.
  • Alan played devil’s advocate, arguing that tenants would likely prefer to retain control over request routing and caching logic rather than giving up control to a third-party HSP (Hosted Service Provider).
  • Alan suggested offering this option while acknowledging that tenants might be reluctant to give up control, and recommended not over-engineering the solution until there’s clear tenant demand.

Energy Architecture Control and Simplicity

  • Glenn and Alan discussed the balance between control and simplicity in designing an energy architecture, particularly focusing on how many “knobs and dials” to expose to users.
  • Glenn argued that minimizing options is better for adoption, drawing parallels to historical debates in computing about providing users with access to infinite resources versus simplified interfaces.
  • The discussion centered on determining the appropriate level of control to give to HSPs (Hyper-Scale Providers) versus hosted workloads, with Glenn suggesting they need to consciously decide who to trust with more control and establish clear boundaries between different parties in the system.

Tenant Control and Delivery Platform

  • Alan and Glenn discussed different approaches to tenant control and delivery in a potential platform.
  • Alan proposed that hyperscalers and CDNs might be early adopters due to their existing infrastructure, while Glenn argued that these entities might prefer more control and be less likely to adopt the platform.
  • They explored the idea of using HSP as a powerful load balancer by having hosted entities provide endpoint IP addresses and bitrate information, which could then be used to optimize delivery across the infrastructure. However, they acknowledged that this model might not support more advanced caching structures emerging from some providers.

Edge Computing Implementation Strategies

  • The team discussed challenges with two models for edge computing implementation: a control-based approach and a helper-based approach.
  • Glenn emphasized that the value proposition lies in providing a bridge between compute platforms and networks, particularly for edge caching workloads.
  • The group agreed to focus on exposing the edge of a specific network rather than trying to harmonize different network architectures across multiple ISPs.
  • Alan announced a demo will be presented at NAB in the SVTA booth, and the team discussed creating messaging materials for the Edge Working Group that could be shared on behalf of the group.

 

Next Meeting:   Thursday, April 16th at 9am PDT.

Have A Question ABout Membership?

Schedule A Meeting

[hfcm id="18"]

Send An Email

Don’t want to schedule a face-to-face meeting just now? No problem. Simply send your membership question to info@streamingvideoalliance.org or fill out the form below and someone will get back to you as soon as possible.