On the release call today I mentioned several takeaways from the OPNFV Euphrates release that might be good to capture as part of the retrospective. Thanks to the release call participants to allow me to take the final 10 minutes for these thoughts!
This release allowed us a pause on project goals that in part at least have shifted to ONAP, and to establish a new/modified strategy for how OPNFV can add value to ONAP and create more synergy in the new networking umbrella structure. The projects include primarily VES and Models, which have rejoined the release program with these goals for Fraser: * VES<https://wiki.opnfv.org/display/ves/Project+Plan>: Test tools and tests for certification of OPNFV reference platform distributions and VNFs as compliant with the VES schema and interoperable with ONAP for lifecycle management based upon VES events. * Supporting that goal is the ongoing work of deploying the Barometer VES agents and the new work of integrating them with ONAP components that will drive lifecycle management (DCAE and microservices that implement closed loop control functions). * The test suite that verifies VES functionality for open source NFV platforms (e.g. the cloud-native platform WIP in Models) and VNFs (WIP in Models) will thus be directly usable by vendors/projects that develop NFV platforms and VNFs, and potentially form part of the Dovetail suite. * This work will enable the Auto project to fulfill some of its goals for platform/VNF assessments under ONAP automation, e.g. per goals for ONAP's role in measuring/managing platform/VNF efficiency, performance, and resilience. * ONAP projects related to VES are more able to focus on the data model and architecture for closed-loop control, and work with OPNFV on the broader goals of promoting VES support in diverse NFV platforms and VNFs * Models<https://wiki.opnfv.org/display/models/Release+Plan>: Use case tests for certification of VIM platforms (cloud-native primarily, but also as time permits OpenStack-based VIM platforms) and VNFs as compatible with VNF orchestration via ONAP for lifecycle management based upon TOSCA-based VNF blueprints. * ONAP has taken over most of the meta-goals of Models e.g. driving collaboration across SDOs and open source projects for modeled VNF support. OPNFV can retain however a key role in developing tools that apply those modeled VNF concepts to a diversity of NFV platforms and VNFs. * Since the overall goal for modeling VNFs is that a single blueprint can be deployed on (and across) diverse NFV platforms, expanding Models focus to cloud-native and hybrid-cloud deployment further fits with the goal of supporting ONAP-managed VNF lifecycle in the true diversity of NFV environments that will be required. This will also be used to support the goals of the Auto project. * Similar with VES, ONAP projects can thus focus on cloud/application orchestration, and benefit from OPNFV's driving the readiness of multi-cloud-compatible orchestration tools (e.g. Cloudify/Aria) for use in the ONAP Multi-VIM project, as applied to diverse NFV platforms. In some cases, e.g. for cloud-native, this work will accelerate ONAP's readiness to expand into hybrid cloud environments in subsequent releases. The other takeaways from Euphrates reflected a shift in strategy for our focus in OPNFV: * Away from (or at least expanding focus beyond) OpenStack-based cloud platforms as the core of OPNFV-inspired NFV platforms. This was partly due to the complexity of the OpenStack platform and the ongoing difficulty in getting OPNFV distros based upon it reliably deployed, but also its insufficiency to meet the goals of evolving beyond first-generation, large-datacenter focused deployments of legacy network functions as VM-based VNFs. Thus the refocus on cloud-native and hybrid-cloud in Models and VES. * Models in particular will produce a variety of cloud-native stacks (Kubernetes, Docker-ce/Moby, Rancher) as test tools, including common subsystems such as SDS (e.g. ceph-docker<https://github.com/att/netarbiter/tree/master/sds/ceph-docker/examples/helm>), monitoring (Prometheus, VES), orchestration (Cloudify). * Related to the above, the insufficiency of OpenStack as currently deployed in OPNFV installers to meet the efficiency and resiliency goals of edge-focused NFV platforms. * We had proposed the OpenStack-Helm (OSH) based Armada<https://wiki.opnfv.org/display/PROJ/Armada> project earlier to address the latter goal (since kubernetes will enable much more resilient management of the OpenStack control plane), but that has not taken off yet. Nonetheless the OSH project is proceeding and getting pushed toward production-readiness by service providers (such as AT&T<http://about.att.com/innovationblog/enterprise_cloud>), and I fully expect there will be renewed opportunity for OPNFV to kick off a project related to OSH in early 2018. * Even so, OSH at the edge may not be efficient or resilient enough, or address the needs of truly lightweight, disaggregated, cloud-native focused VNFs. So we have refocused our efforts to establish the same common ONAP support goals for modeled VNFs and closed-loop control toward those things in cloud-native platforms. Thanks, Bryan Sullivan | AT&T
_______________________________________________ opnfv-tech-discuss mailing list opnfv-tech-discuss@lists.opnfv.org https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss