Attendees:
Please add your name in here:
- Mark Shostak (AT&T)
- Karine Sevilla (Orange)
- Xavier Grall (Orange)
- Trevor Cooper (Intel)
- Al Morton (AT&T)
Agenda:
- Agenda Bashing
- Discuss network test metrics tables w/ Xavier; identify what is Infra generated, what is Instrument generated and move Instrument metrics to RMCH-08 (Xavier)
- Clarify and define capabilities/metrics/constraints vs. resources/capabilities, then normalize chapter
- Clarify CPU overcooking description
- Discuss L3 cache management (Sridhar)
- Perform live review for end-to-end chapters 4 and 5
- Issues
- Other Issues: List of RM Chapter 4 Issues
- Other Issues: List of RM Chapter 5 Issues
Actions:
- Issue #377: Create Issue to document and define the term "Measurement", and replace instances of metric(s) with measurement(s) in RMCH-4 (PENDING REVIEW (PR #386))
- Issue #378: Create Issue to document and remove measurements made by external test instruments in RMCH-4
- Issue #379: Create Issue to document and add measurements made by external test instruments to RMCH-8
- Issue #380: Create Issue to document and define the term "Performance Measurement", and replace instances of monitoring with performance measurement in RMCH-4 (PENDING REVIEW (PR #386))
- Issue #381: Create Issue to document and define the term "Monitoring", and apply to RMCH-4, as appropriate (PENDING REVIEW (PR #386))
- Create corresponding Issue(s) for any of the above RMCH-4 Issues, that apply to RMCH-5
- <CLOSED> Create/Update next week's agenda
Minutes:
- Lengthy discussion of the network test "metrics" tables and various terms/definitions was held, resulting in the following conclusions:
- The term "metric(s)", as applied across the chapters, was ambiguous in that it did not explicitly nor implicitly denote a measurement. It was proposed and unanimously agreed that the term metric would be replaced with measurement, to make it clear the intent of the document in said cases is to refer to a measurement, as opposed to a specification
- Measurements in the chapter are intended to be measurements that can be made/taken by the infrastructure, without any external instruments. For example, measurements taken by kernel tools, or by hardware integrated with the hosts (e.g., packet counters). Multiple instances of measurements that were taken by external instrumentation are currently found in the chapter. It was proposed and unanimously agreed that all measurements requiring external instruments, be moved to chapter 8 (Verification and Validation), where infrastructure performance profiling is performed
- The term "Monitoring" [capabilities] is currently used to describe and refer to measurements (formerly known as metrics), which are taken internally by the infrastructure itself. It was proposed and unanimously agreed that all measurements which can be taken internally by the infra, be renamed "Performance Measurement(s)", aka PMs
- With the term "Monitoring" being vacated by the decision in the previous bullet, it was proposed and unanimously agreed that monitoring be defined so as to refer to passive observation functionality. For example, the ability to sniff a packet stream, al la tcpdump, Ethereal, etc.
- An agenda item for the next meeting is to identify and document Performance Measurements (PMs) to add to RMCH-4, replacing the measurements relocated to Chapter 8
- Thanks to the team for making it through these deliberations!