============================================== Minutes of the IPFIX session at IETF 73 Minneapolis Wednesday November 19 at 0900-1130 Taken by Rolf Wolter ============================================== Chairs: Nevil Brownlee Juergen Quittek 0) Session summary 1) Agenda review and WG Status 2) Internet-draft Status 3) Information Elements Registry Status 4) Drafts from old charter 5) Drafts from the new charter 6) Milestone review 7) AOB 0) Session Summary Seven documentis in the RFC editor queue are waiting for the PSAMP info model: four IPFIX documents (architecture, applicability statement, reducing redundancy and testing guidelines) and three PSAMP documents (framework, packet selection, protocol) Now the PSAMP info model has entered the RFC editor queu and all eight documents can hopefully be published soon. Four further IPFIX documents have passed WGLC. Three of them (file format, export of type information, MIB module) have all issues cleared and publication as RFC will be requested for them by the chairs after the session. The fourth document, IPFIX over a single SCTP stream, has a sigle issue with a normative reference to be solved. Work in the WG focuses on the two remaining topics in the charter: IPFIX configuration and IPFIX mediation. for both topics, WG IDs exist, they need further work. 1) Agenda review and WG Status There were no requests to modifythe agenda. 2) Internet-draft Status The status of WG IDs was described as in the session summary above. 3) Information Elements Registry Status Nevil: IANA is working on an XML page for this. It will have the IEs from RFC 5102 and draft-aitken-ipfix-new-infos-01.txt. Working with IANA took quite a while, finally we have all information in XML file published in the registry. This will become the primary source for publishing information elements. IPFIX was the first candidate for this process. Finally things are looking good! 4) Drafts from old charter IPFIX MIB - draft-dietz-ipfix-mib-05 ========================== Benoit Claise: Version 5 has been published, all issues solved and the ID is ready for next steps. Add ons: - transport session mode distinguishes between Exporter and Collector (the MIB is used for both) - template can now specify Enterprise Elements 5) Drafts from the new charter IPFIX Per-Stream SCTP reporting - draft-ietf-ipfix-export-per-sctp-stream-00 ============================================ Benoit: WGLC finished. Clarifications on the interaction with RFC5101. Only one opern issues left: There is a normative reference to SCTP-RESET draft (draft-stewart-tsvwg-sctpstrrst-00.txt) from the Transport WG. This ID describes allows the addition of streams within an existing SCTP association as this functionality is required. Dan (AD): This draft cannot be a normative reference as long as it is not adopted by a the TSVWG. Suggestion: describe the mechanism in the normative part of the text and add an informative reference to the SCTP-RESET document. Once the document is included in the Transport WG charter it can be used as a normative reference. Juergen: Authors of draft-stewart-tsvwg-sctpstrrst-00.txt should review the inserted text. IPFIX File Format - draft-ietf-ipfix-file-03 ======================== Brian Trammell: Title changed to "Specification of the ipfix file format". Went into WGLC after Dublin, received good feedback: - added flow start/flow end IEs for timestamp resolution - guidelines for message boundaries Next steps: submission to IESG now Exporting Type Information for IPFIX IEs draft-ietf-ipfix-exporting-type-02 ================================== Brian: Status: completed after Dublin meeting. Configuration Data Model for IPFIX and PSAMP - draft-ietf-ipfix-configuration-model ====================================== Benoit (on behalf of Gerhard Muenz): Lesson learned: modeling of metering concepts sounds simple but are quite complex when taking various corner cases into account! Selection process was changed to PSAMP definition. Cache layout description modified to deal with issue that an information element is not present. All flow key fields must be included in the flow unless it cannot be deduced. New parameter for exporting process added: rate limit TLS versus DTLS: should it be configured or not? No consensus achieved on the mailing list, the proposal is not to include Brian: Deep configuration of TLS parameter would increase complexity, however identify is required. Proposal: split in two, Brian will draft some text. Benoit: YANG dependent versus independent part: possible split the two parts. Yang draft expected in October 2009. Dave Harrington (co-chair of netmod): to be submitted in September 2009. Benoit: What shall we do? Jürgen: Avoid further dependencies Dave Harrington: it would be better to be consistent Benoit: keep as single draft and split later, if necessary Jürgen: Single draft with structure for two parts: YANG dependent and independent part. IPFIX Mediation: Problem Statement - draft-ietf-ipfix-mediators-problem-statement-01 ================================================= Atsushi Kobayashi: Background described, started at IETF69 New milestone: submit to IESG in April 2009 Summary of ipfix mediation, it consists of a set of functions: - Rerouting input data to an appropriate Collector - Replicating input data - Filtering and selecting input data - Aggregating input data - Modifying input data - Changing a transport protocol carries IPFIX Messages - Correlating a set of Flow Records/Packet Reports and calculating new metrics. IPFIX Concentrator can be utilized to reduce traffic flowing toward a central collector by aggregating flows. Correlation of Flow Records creates new metrics: one-way delay, jitter Aggregation types: - Space composition: aggregation of Flow Records involved in a larger Observation Domain or a set of Observation Points, e.g. link or domain aggregation - Time composition: aggregating consecutive records within the same flow Next step: - include feedback received from Gerhard and Nevil - clarify the focus of this document - can mediation be implemented at any ipfix device? Benoit: We should not include all 3 drafts but don't go into the details from the three drafts. The drafts should be included in the overall framework. IPFIX can be implemented everywhere but it requires logical functional block. Nevil: I agree that the existing drafts should build on the framework. Suggestion: have short summary of the content of these 3 drafts. Juergen: Describe what are the features and why should it be included Brian: Full agreement with Nevil's proposal and Benoit's answer about implementing mediation. Provide a terminology and reference for implementation. Christoph via jabber: Definitions should be moved to framework draft, add more details to avoid confusion. Juergen: agreed Chris: Mediation creates a holistic view for flows from different exporters, bring information from multiple flows into one combined flow. This problem hasn't been addressed in the flow yet - holistic flow. So far the document addresses scalability issues, the holistic view (trace flows as they travel through the network) is another aspect of mediation and should be included. Nevil, Juergen: This should be added IPFIX Mediation: Framework - draft-ietf-ipfix-mediators-framework-00 ========================================= Atsushi Kobayashi: History: WG draft in IETF71 approved Various improvements included New Milestone: June 2009, submit to IESG Details: - Flow expiration: aggregation needs similar condition than the metering process - Reuse IM from RFC5102 Next steps: - ensure text consistency with ipfix and psamp docs, terminology, mediation functionality - submit next version in Feb 2009 Juergen: This document is intended to be informational. Should we have IEs in this documents? Probably not, however there are no other documents yet Brian: IEs shouldn't be here, this requires a new document. Juergen: Protocol specific issues for implementation should be included in a separate standard track document that identifies these. Specify scope of the document, e.g. what additional pieces are required. Juergen: Call for action, read the two mediation documents and share comments on mailing list as this is the next big step for IPFIX. What can we learn from other mediation projects, e.g. mediation for accounting and billing? Mediation for accounting is different from flow mediation, still it would be good to have some input. Juergen will post some references to the mailing list. 6) Milestone review Done: Submit File Format I-D to IESG for publication as Standards track RFC Submit IPFIX MIB draft to IESG for publication as Standards track RFC Submit Type Export I-D to IESG for publication as Standards track RFC - Jan 2009: Submit Single SCTP Stream I-D to IESG for publication as Inform. RFC Nevil: is this realistic? Benoit: yes, leave the time frame as only minor changes are required - March 2009: Submit Configuration Data Model I-D to IESG for publication as Std. track RFC - April 2009: Submit Mediation Problem Statement I-D to IESG for publication as Inform. RFC - Jun 2009: Submit Mediation problem statement I-D to IESG for publication as Inform. RFC - tbd 2009: Publish Configuration Information Model I-D Publish Configuration Yang Data Model I-D -> as agreed today we will keep both parts together in one document as long as possible 7) AOB Dan Romascanu: Could we have a summary from the NMRG workshop? Juergen: Reporting flow metrics, e.g. exporting IPPM metrics Brian: Discussions at ippm yesterday, e.g. use IPFIX data to derive passive performance measurements. Potentially combine some work from the two WG Al Morton: New IPPM work will be prioritized now, so if passive measurement has a priority in ipfix it could also be used there Tanja: new version of the "flow selection draft" will be presented at the next meeting Juergen: Anonymization was another topic at the workshop, it's needed but more a research issue than a matter of standardization Brian: Ground work already going on in this area, a draft might be ready for the next meeting. Benoit: Topics such as SQL data base, how to collect 10GB in hardware/software were discussed, too early for standardization. Brian: IPFIX tutorial given at IMC.