Sorry, you need to enable JavaScript to visit this website.

Feedback

Your feedback is important to keep improving our website and offer you a more reliable experience.

2019 StarlingX Deep Dive Meetup in Beijing

BY 01 Staff (not verified) ON Oct 29, 2019

Intel aligned with 99Cloud, WindRiver and FiberHome to co-organize the "2019 StarlingX DEEP DIVE MEETUP" in Beijing, Sept. 20, 2019. More than 80 participants from 22 companies participated including all major PRC telco operators, Cloud Service Providers and edge computing solution providers. Experts from Intel, China Unicom, FiberHome, Wind River, 99Cloud, China UnionPay and China Mobile shared use cases of edge computing and experiences deploying StarlingX. A demo showcase and enthusiastic group discussion gave participants additional opportunities for direct communication to share requests and feedback.

Use Case Study

The meetup began with a joint presentation by WindRiver, Intel, Fiberhome and 99Cloud, which have contributed the majority of StarlingX. This introduction painted a whole picture from multiple dimensions, from project history to today’s community ecosystem, use scenarios to technical evolution on architecture, and from current 2.0 feature highlights to the outlook in future releases.

Presentations

Several major partners/customers came to the stage and shared their solutions or product practices with StarlingX as the infrastructure:

China Unicom Cloud Division: StarlingX Practice in Industrial Manufacturing Edge Cloud

Besides as the traditional operator, China Unicom is also pursuing CSP business opportunities on market segments such as industrial manufacturing edge, where edge computing platform holds the valuable position, connecting IOT and industrial fields with cloud centers. This presentation proposed a blueprint and explained how IaaS, PaaS and SaaS should be laid up layer by layer for fulfilling industrial needs. StarlingX is a favorable solution for IaaS, benefitting from dedicated optimizations for edge such as low latency, real-time, security, and high availability, in addition to orchestration capabilities inherent in Kubernetes and OpenStack. China Unicom articulated how StarlingX distributed-cloud feature can be used to deploy cross-regional edge platform in industrial cases.

FiberHome: Multi-cloud and Edge Platform Convergence

FiberHome has developed a full product portfolio on Cloud (FitCloud) and Edge (FitEC), and various of cloud segments: public and private cloud to hybrid and industrial edge. To improve technology competency, innovation, and resource utilization/efficiency, FiberHome moves ahead in the direction of multi-cloud and edge platform convergence. StarlingX has been selected as Edge infrastructure, after FiberHome evaluated several solutions available in open source communities. As a result, FiberHome enrolled as a StarlingX community core member (with TSC representative) and has committed to make continuous contributions to StarlingX project since 2018.

99Cloud: Edge Platform Application on Smart Agriculture.

This case drew audience interest not only for its vivid example (cultivating shrimp), but more for the challenges of landing the edge platform in fields (around shrimp ponds) and how 99Cloud took advantage of StarlingX capabilities to address those them. In the real deployment, the excellent reliability and high scalability (from an All-in-One Simplex to dozens of hosts managed by distributed-cloud feature) were explicitly highlighted and appreciated by customers.

China Unicom Research Institute: MEC (Multi-Access Edge Computing) Platform and Implementation

MEC was driven by 3GPP and ETSI. Following ETSI MEC framework and architecture, China Unicom proposed their reference design on layers, from OSS/BSS, MEAO/NFVO, VNFM down to MEPM/MEP. And practically in order to fulfill the requirements on high bandwidth, low latency and huge scale of connections, China Unicom Edge Business Platform choose StarlingX as its virtualization layer infrastructure solution. Besides the dedicated characteristics designed for Edge, the full compatibility with ETSI NFV also directly helped StarlingX win a position in China Unicom Edge Platform.

China UnionPay: Research of Edge Computing Empowering Financial Applications

By nature, financial applications (for instance, payment use cases) have the special requirements, such as security and privacy, dynamic and fragmented workloads, and so on. In this session, China UnionPay as the official designated vendor to Chinese banks, shared their vision and technical approaches how they would utilize edge computing platform to empower typical financial applications. For instance, very recently China UnionPay and Intel have been team-up and developing a POC about the contactless payment system for electronic-vehicle charging stations, in which StarlingX servers work as the secure and scalable edge platform to host video surveillance workloads, charge metering and secure payment modules. More details will be exposed in incoming Open Infra Summit 2019 November Shanghai. In the context of deploying financial applications on edge, among all existing open-sourced projects, StarlingX is having lead position for sure, benefiting from the hardened security design (specifically with well integrated TPM framework) and backed by security technologies like Intel SGX on Intel platforms.

China Mobile Research Institute: Heterogeneous and Collaborative Management Requirements on Cloud and Edge

other carriers, China Mobile is also aligning their edge platform with ESTI MEC framework and reference architecture, though there are certain level of differences in implementation details. For example, in China Mobile’s Edge solution, ECM (Edge Cloud Management Center) and ECMP (Edge Cloud Management Platform) together serve as MEC MEO (System Level ME Orchestration), ECP (Edge Cloud Platform, PaaS) covers both MEPM (host level ME Platform Management) and MEP (ME Platform), and ECI (Edge Cloud Infra) matches with NFVI. Facing the natural complexity of edge cloud, in terms of multi-locations (cross regions), heterogeneousness, collaboration, and autonomy, China Mobile shared the experience learned from their practice. Regarind to the relevance to StarlingX, China mobile placed StarlingX as the NVFI to cover infrastructure capabilities originally covered by OpenStack or/and Kubernetes.

In summary, These 6 case sharing above covered several typical scenarios on edge-computing. On the big picture, no doubt StarlingX can play an important role on all these designs, and indeed it is having a competitive value proposition with highlighted features dedicated for edge computing, such as low latency, high availability and harden security. However, going further with deeper discussion and analysis (carried by the panel discussion groups at afternoon), we have collected tons of valuable recommendations, honest feedback and constructive comments on almost all aspects, like less footprint or system overhead, better usability (easy-to-use), more diverse community, more production landing or joint-projects with other communities, and so on, just name a few. We will the output from panel discussions in details, as follows.

Demo Showcase

Mr. Hu Wei from Intel introduced the edge computing demo application of StarlingX in different stages and scenarios to the participants, and brought StarlingX cooperation and joint demo projects with ecosystem partners. He said that the StarlingX community is looking forward to more partners to join, to accelerate product development through forward-looking interoperation and cooperation.

AI Face Recognition Applications on StarlingX Edge Clouds (to showcase Automated Deployment, Data Acquisition, and Model Upgrade usage)

In this demo case, the central cloud performs complex model training and then pushes the model to the edge cloud. In addition, StarlingX 2.0 combines Kubernetes and containerized OpenStack to take advantage of the containerized deployment and enhance platform delivery, which is more convenient to mission critical application scenarios. Based on this architecture, a face recognition application is automatically deployed on the edge cloud. At the same time, combined with the container as a service (CaaS) platform, the user is provided with a development environment, the machine learning module can be uploaded to this, and the system is checked and updated to StarlingX to implement the model upgrade.

Edge Workload Integration for Linked Cloud Services Based on ACRN and StarlingX. In this demo case, the end user can remotely control the edge devices for lifecycle management and collect real-time data of the motor to the edge server side. A containerized service with Data Analytics Reference Stack (DARS) is deployed on top of StarlingX for motor data analysis and anomaly detection. Through StarlingX's orchestration, edge devices can support Clear Linux, Preempt-RT Yocto and Windows systems as virtual machines on top of the ACRN hypervisor, which is a Type 1 hypervisor that runs directly on bare metal hardware and is suitable for a wide range of IoT and embedded device solutions.

Android in Container (AIC) based on StarlingX 

The AIC scenario demonstrates the implementation of Android as a Guest OS in the Kata secure container and fits the application scenario of StarlingX from the cloud to the edge. A secure, lightweight "virtual machine container" solution from the Kata Secure Container, complemented by the StarlingX Edge Platform, is the perfect combination with Android to bring a faster, more secure solution to the industry.

Deploy and apply TSN applications in StarlingX 

TSN is a set of standards defined by the IEEE to support the transmission of time-sensitive data over Ethernet. It is one of the key technologies in many areas of edge computing such as autonomous driving and industrial. This demo shows how to deploy a TSN application on top of StarlingX and showcases the effect of transmitting time-sensitive video streams using TSN technology based on the Intel 1210 NIC.

Also, there are other use cases like Big Data, Robotics and etc. They are introduced orally and not brought to onsite demo booth. Wei also introduced the concept of the demo lab which facilitates StarlingX members to get access to those demo projects for further evaluation and interoperation.

In conclusion, StarlingX is a fully open sourced, industry leading software platform, optimized for edge computing and designed for edge deployments. We expect to bring forward more use case demo and cooperation with our ecosystem partners in the future.

Group Discussion

  • Documentation still has some gaps about practical HOW-TO steps and design documentations (arch, call flow and APIs) on core components (e.g. flock services).
  • User-friendly and developer-friendly are crucial to make an open-source project prosper. (for instance, Chinese users have difficulty to access k8s or docker registry, can we setup one registry to ease StarlingX installation??) 3. Can we extend StarlingX a bit more than IaaS? Having some PaaS capabilities will help industrial customer/production adoption.
  • Security (including data privacy) is important or even critical for industrial scenarios, what can be done make it one more differentials of StarlingX
  • less footprint, supporting to smaller nodes and manageability on on-premise (or legacy hosts), mandatory to IOT cases.

StarlingX in Telco-group1

  • StarlingX deploys scripts and documents to Centos.K8S needs to overturn the wall when pulling resources. This is unfriendly for the environment that cannot be overturned. It is expected to have a guiding plan to address domestic solutions.
  • Source comments are too simple, expect to optimize.
  • The core project project architecture diagram, communication, call relationship architecture document, technical framework document used.
  • How to start each individual project and how to publish it.

StarlingX in Telco-group2

  • Building a local registry, and providing an installed version of the minimum feature set which is convenient for initial user evaluation.
  • The performance test framework feature will be integrated and reinforced by incoming StarlingX3.0 to meet the evaluation requirement of the mission critical use case.
  • The vitality of open source organizations is less cultivated in the superb technology than in the friendliness to beginners. This was an insightful perspective that we could count on to define key performance indicators for community development work, to achieve the objective and key results of the growth hacker.

StarlingX in  Industry & IOT group

  • About making StarlingX pass CNCF certification, is there such a requirement or expectation from industrial users ?
  • Besides establishing a new StarlingX based edge cluster, industrial users do expect StarlingX is also capable to manage legacy hosts, which they have been investing for years.
  • Starling or Kube-edge, under what circumstance, one solution works better than another in IOT user cases? Specifically, StarlingX mainly fits edge server fragment, while Kube-Edge does better on far-end edge devices.
  • Expect StarlingX to have/enable more components (protocols) or services, so in certain sense, it is extending to PaaS.
  • IOT users do care about security, data privacy and regulations. StarlingX has been doing good jobs on this area, though, we need do better job to make users aware of the achievements.
  • Expect StarlingX to support smaller nodes (including X86 PC and ARM hosts) as computing nodes.
  • Suggest StarlingX to show the minimal requirements on Hardware spec and make a roadmap to optimize the footprint (CPU, memory and storage). Can we start this from 4.0 ?
  • What's the status of multi-tenancy, on K8S and OpenStack in StarlingX ?
  • Look forward to seeing more use cases or blueprints which demonstrate the integration of solutions from multiple community projects.
  • Suggest to make "system application-apply" support specifying different namespaces.
  • Expect to see multiple layers of logging, monitoring and metering, on different types of resources: HW, OS, containers, VMs, and so on.
  • What is the strategy of upgrading Kubernetes?

Special Thanks

Special thanks Hu Wei, Hu Yong, Wang Haitao, Wang Yu, Liu Yuan, Wu Xiaowei, Qi Mingyuan, Shane Wang, Ding Jianfeng, Yang Ailin, Jin Yuntong, Suzie Yang for planning, organizing and engagement this event. Your solid work and engagement made this event successfully launch. Special thanks to Cindy Xie for delivering Keynotes, Maggie Liang for guiding on the whole event organizing. All of your effort demonstrated Intel leadership in Open Source community.