[Edge-computing] Open Infrastructure Summit and PTG Shanghai - edge recap
ildiko at openstack.org
Mon Dec 2 15:13:51 UTC 2019
Hereby is a recap of the Open Infrastructure Summit that was held in Shanghai, November 4-6, with conference sessions including presentations and panel discussions, a lively Marketplace and the co-located Forum to provide space and time for users, operators and developers to talk about the software pieces the communities are working on.
As usual the event started with keynotes that were focusing on open infrastructure and updates about OpenStack and further OSF projects as well as news from the global and local communities. During the event the sessions were organized into dedicated tracks to cover relevant areas including a combined track for NFV, 5G, and edge computing. OSF Edge Computing Group contributors had a panel discussion to give an update on the latest activities of the group including a sync with representatives of the Chinese edge community.
Similarly to former Summits we had the Forum running in parallel to the conference with 40 minutes long sessions with users, operators and developers in the rooms to discuss feedback, new ideas and next steps. StarlingX and the OSF Edge Computing Group had a joint session to talk about reference architectures and gather new use cases from the attendees. The session discussed use cases related to Content Delivery Networks (CDN) and requirements to run workloads on bare metal at the edge and further items in the area of networking.
The conference sessions were recorded and the slides and videos are being uploaded to the event website as you read this article.
We had a full day dedicated to the edge working group to talk about recent activities, synchronize with OpenStack and OSF projects and work on the roadmap for the first half of 2020.
The working group discussions were mainly focusing on testing activities that we have started earlier this year. The group is working on defining reference architectures that we are turning into reference implementations to see if they fulfill edge requirements and find gaps and areas to improve. In recent activities we started to test the distributed control plane model with using StarlingX and setup the centralized control plane model with TripleO.
Due to limitations on the lab environment we have we hit a roadblock with the latter scenario which brought us to discuss alternatives both on the side of ways to deploy the test systems as well as alternatives for lab environments. The group decided to reach out to universities as well as look into offers on lab donations from earlier PTGs.
The group has a test plan proposal that we started to review at the PTG. Our main goal for testing is to evaluate the functionality of the services that we deploy in the lab environments based on information we gathered from collecting use cases from different industry segments. Evaluating the services as a complete platform is out of the scope of our work as the group is not working on an end-to-end solution that can be used out of the box but rather wants to make sure that the building blocks deliver what they promise. In that sense we will not test the steps of deployment and configuration neither the steps of upgrading the infrastructure at this stage.
Based on the outreach to look for lab environments we also hope to be able to deploy edge environments at multiple sites which could give us the possibility to eventually test geographically distributed configurations. This is not a high priority task at this point, but the group had an agreement to keep this as a desirable goal.
After the test environments we also briefly discussed possible workloads to test our edge environments with. We identified two areas to look at, these being vRAN and cloud gaming. Both of these use cases are already moving to deployment stage which means that we have a potential to find open source building blocks to use or partner up with organizations to do joint work in the area, one example to look at is O-RAN.
The group is planning to record demos once we reached the stage of operational deployments with workloads running on top of them. We are also still dedicated to lower level testing such as Keystone federation testing in a Devstack environment.
Another big chunk of the PTG activities of the group were cross-project discussions with two main objectives. As projects and services continuously evolve we need to periodically re-evaluate our reference architecture models to ensure they are still correct and realistic when it comes to the details. We identified items to fix for projects like Glance, Nova and Cinder. Beyond synchronizing with the projects we’ve already been working with we are also constantly looking into new projects to add to the models and activities as well. Candidates high on our list are Ironic and Cyborg as both managing bare metal resources and hardware acceleration technologies are getting more and more crucial for edge use cases.
We had sessions with Cinder, Cyborg and Neutron and had one large joint session involving Ironic, TripleO, Airship and Manila.
With Cinder we focused mainly on deployment options regarding the placement of the services and some functionality around availability zones (AZs). We touch based on features such as setting a default volume type per AZ as volume requests don’t contain AZ information. Another feature that is very useful for edge use cases is the instance locality filter which can ensure for single node edge sites that the deployment is using local disks on the node.
Group members had a very fruitful discussion with the Cyborg team to kick off the collaboration between the project and the edge working group. As hardware acceleration is a hot topic for edge use cases it is very important that we have the possibility to experiment with that in our test environments as well beyond capturing this need on architecture diagrams. During the session we synced up on the state of both Cyborg as well as the work of the Edge Computing Group and started to explore some relevant use cases. Our current focus is on vRAN deployments, and mentioned the work that a team from Rakuten was presenting at the Summit.
We had a short sync with the Neutron team to check on items that are in progress. Our focus with that team was on enhancements in the ability of managing network segment ranges in edge deployments. The task is currently approved and in progress with some cross-project work items in the plan such as collaboration with the Placement team. We will track progress during the Ussuri release cycle and include the functionality in our testing activities once it is implemented.
Our last session was held with a large group where we had members from multiple project teams. Our main focus was concentrating around Ironic use cases mainly around the area of virtual media based provisioning as well as booting from a URL. For the deployment side in relation to TripleO we were discussing an option for single step configuration that would include full node settings for the hardware. It faces challenges around locality to ensure that the image has a local source or is cached somewhere close to the server that is getting deployed. To utilize Ironic for Airship we mentioned the Metal3 CRD work. The last item of the session was a short update from Manila to summarize the relevant service and ongoing work around active-active configurations.
Based on the information the working group has gathered during the Open Infrastructure Summit and PTG we will work on both building and evolving test environments as well as to enhance our reference architecture models. The group is also planning to start actively discussing concepts such as containerization and extend the architecture models to be suitable for all types of workloads.
You can also check out the StarlingX blog for details about the StarlingX activities in Shanghai.
For more details about the OSF Edge Computing Group check out the group's wiki page. You can participate in the discussions by joining our mailing list or joining the #edge-computing-group IRC channel on Freenode. The group will also participate in the next OSF event in June, 2020. Stay tuned for updates!
Please let me know if you have any questions.
More information about the Edge-computing