Cloud Native + Open Source Virtual Summit China 2020

Based on Kubernetes, providing unified management of heterogeneous computing, network, and storage resources;
Leveraging GPU virtualization to build a GPU shared resource pool to achieve more fine-grained quota management of GPU resources、multi-tenant sharing scheduling of a single GPU、maximize GPU resource utilization and improve resource efficiency;
Implement a custom Kubernetes scheduler for resource scheduling management of GPU tasks according to priority;
GPU visual monitoring based on Prometheus with the ability to aggregate performance metrics by any dimension and the monitoring granularity can up to a single card level;
One-stop scientist workbench through Kubernetes scheduling mainstream artificial intelligence framework, algorithm scientists complete end-to-end algorithm development iterations, model training and release of artificial intelligence services in the cloud.

Monoliths! Who talks about Monoliths anymore? Didn't we move on to SOA and microservices a decade ago?

Despite the domination of distributed architectures for cloud software, 60% of enterprise IT budgets still go to the datacenter, where the Monolith is King. While these legacy applications have historically been perceived as liabilities, new developments in service mesh enable enterprises to benefit from their monolithic assets while taking advantage of Cloud Native technologies.

In this talk, you will learn about how you could embrace your monolith by enabling it with:
* Zero Touch security with Envoy and Spiffe
* Zero Touch observability with Prometheus
* Zero Touch uptime improvements with Circuit Breakers
* Zero Risk test in production with fault injection and canary deployment
* Zero Risk traffic migration in Hybrid cloud environments (VM and multi-cluster)

With the scale and complexity of a Kubernetes cluster increasing, it's getting harder to keep the cluster deliver qualified pods with high success rate and low latency. In this session, engineers from Ant Financial would share their experience on design of SLO architecture and methods implemented to achieve high service level objectives. They would introduce proper indicators proposed to measure whether Kubernetes clusters are healthy firstly. Then, they would explain how to design and implement the tracing and analysis platform to collect valid metrics and compute those indicators. With the tracing and analysis platform, problems occurred in pods' delivery process would be diagnosed easily. Finally, they will show how to precipitate artificial experience to self-healing system to fix known problems automatically.

How SPD Bank reacted to new business transformation，utilized K8s to establish PaaS platform specific for financial industry，converted core business services with long-standing issues into container based infrastructure，dealt with performance challenge of Double 11 successfully. And how PaaS platform scaled to satisfy increasing requirements and challenges after core business system with huge traffic going.

1.Comtruct PaaS platform specific for financial industry based on K8s which holds up nearly 100 core business systems and met the requirement on nearly 100% continuity of key services and supported peak time transition of Double 11 promotion for SPDB with historical value of TPS
2.how the platform supports multiple A Level applications
3.how to builds a cross-regional multi-cluster networking solution based on CNI network
4.Support the dual-stack operation requirements of IPv4&v6

Meituan-Dianping is a world leading local life service platform. Migrating cloud infrastructure like it to Kubernetes is not easy. This road is complex and difficult, but it’s all worth it.

In this talk, they will share the lessons learned from their work in the past two years from a technical perspective, including:

1. What are the main obstacles and benefits to large-scale cloud infrastructure migration from OpenStack to Kubernetes?

2. What’s the future of VM? How to avoid maintaining two cloud infrastructure systems? All(containers and VMs) run on Kubernetes.

3. How do we deal with the problems of reliability and efficiency brought by large-scale Kubernetes cluster and dozens of clusters?

4. What are the main technical flaws that hinder cloud native? How to do that if your application management is different from Kubernetes and you have to rely on your existing infrastructure?