Overview

With the rapid development of cloud native technologies, many applications are going cloud native. From 2021 to 2022, the total number of cloud native applications in Kubernetes clusters has increased by more than 30% year-on-year. Kubernetes is becoming the platform for running almost anything — a virtual "operating system" for cloud native applications in the cloud era. However, further research shows that the CPU usage of most user nodes in Kubernetes clusters is less than 15%. According to a survey conducted on a range of clients, the primary causes of low resource utilization may be summed up as follows, if interference elements like idle resources and package activities are taken out:

Nodes are deployed in different clusters. They cannot share compute resources with each other, resulting in an increase in resource fragments.
The node specifications are not ideal for applications that undergo frequent changes. At first, the node specifications match the application requirements, resulting in a high resource allocation rate. However, as the applications evolve, their resource demands change, causing a significant difference in the ratio of requested resources to node specifications. This leads to a decrease in the allocation rate of node resources and an increase in compute resource fragmentation.
There are a large number of reserved resources. Online services experience daily peaks and troughs. To ensure service performance and stability, users apply for resources based on peak usage, which may result in many idle resources in the cluster during certain times.
Online and offline services are deployed in separate Kubernetes clusters, and resources cannot be shared between them at different time. This means that during off-peak hours for online services, the resources cannot be used by offline services.

These are typical instances of the difficulties encountered when creating cloud native applications. Various deployment solutions are needed for different service architectures during cloud native progress. Applications with varying architectures evolve at different paces, so development teams must balance service performance with service quality. How can these complex scenarios be made simpler so that customers can gradually increase resource usage and reduce costs?

CCE has created a cloud native hybrid deployment solution based on the Volcano and Kubernetes ecosystems, which helps users improve resource utilization, reduce costs, and increase efficiency. This solution is the result of years of exploration and practice in hybrid deployment. Hybrid deployment goes beyond running multiple services in a single cluster. It ensures that user applications are deployed in optimal locations and that the required resources are available. This is the core principle of cloud-native hybrid deployment: hierarchical resource management.

Hierarchical Resource Management

Hierarchical resource management ensures that applications can access to the necessary resources in their designated environments after they have been scheduled.

Resources are isolated from multiple dimensions like CPU, L3 cache, memory, network, and storage based on Huawei Cloud EulerOS 2.0. Additionally, resources, mainly in the kernel mode, with some in the user mode, can be quickly preempted in milliseconds or evicted in seconds to ensure the quality of online services.

Resource isolation measures (such as CPU pinning, NUMA affinity, tidal affinity, and network bandwidth control) ensure the resource-sensitive services to meet their SLOs.
Resource priority control (such as hierarchical CPU suppression, memory tiering, network priority control, and disk I/O priority control) improves resource allocation and has little or no impact on the SLOs of high-priority services.

Hierarchical resource management provides a basis for deploying online services and hybrid deployment of online and offline services. This resolves problems that a large number of resources are reserved for applications and resources cannot be reused at different times.

Online and Offline Jobs

Jobs can be classified into online jobs and offline jobs based on whether services are always online.

Online job: Such jobs run for a long time, with regular traffic surges, tidal resource requests, and high requirements on SLA, such as advertising and e-commerce services.
Offline jobs: Such jobs run for a short time, have high computing requirements, and can tolerate high latency, such as AI and big data services.

Features

Function	Description	Documentation
Dynamic resource oversubscription	Based on the types of online and offline jobs, Volcano scheduling is used to utilize the resources that are requested but not used in the cluster (the difference between the number of requested resources and the number of used resources) for resource oversubscription and hybrid deployment to improve cluster resource utilization.	Dynamic Resource Oversubscription
CPU Burst	CPU Burst is an elastic traffic limiting mechanism that allows temporarily exceeding the CPU limit to reduce the long-tail response time of services and improve the quality of latency-sensitive services.	CPU Burst
Egress network bandwidth guarantee	The egress network bandwidth used by online and offline services is balanced to ensure sufficient network bandwidth for online services.	Guaranteed Egress Network Bandwidth