Author: Hai Er
Trusted Cloud Evaluation is a professional evaluation system for cloud computing services and software under the China Academy of Information and Communications Technology. After ten years of development since 2013, the Trusted Cloud Service Evaluation System has become increasingly mature and has become a government-backed, industry-standard, and user-selected system. An important reference for the type.
In May 2022, the State-owned Assets Supervision and Administration Commission of the State Council formulated and issued the "Work Plan for Improving the Quality of Listed Companies Controlled by Central Enterprises", focusing on the environment, society and corporate governance (ESG), and put forward requirements to "implement the new development concept and explore the establishment of a sound ESG system". FinOps, GreenOps and ESG concepts are highly compatible. Cloud computing is the engine and cornerstone of enterprise digital transformation. Cloud computing-oriented cost optimization and low-carbon operations will help companies practice ESG development concepts and fulfill social governance responsibilities.
In this context, China Academy of Information and Communications Technology officially launched the "Cost Reduction and Efficiency Increase, Green Development 丨 Cloud Cost Optimization and Low-Carbon Operation Pioneer Practitioner" enterprise selection activity, and established a new paradigm of cloud cost optimization and low-carbon operation through the selection of benchmark cases , Pioneering enterprises under new technologies, helping cloud cost optimization and low-carbon operation concepts continue to infiltrate and put into practice in enterprises, improve the value of enterprises using the cloud, and promote the green and low-carbon development of the cloud computing industry.
In this selection of pioneer practitioners of FinOps cost optimization, Jikr Auto won this honor based on the active practice of Alibaba Cloud's FinOps cost management system in the past year. Through the ACK One multi-cluster management and cloud-native resource portrait functions provided by Alibaba Cloud Container Service, it checks the resource usage of dozens of online and offline K8s clusters and thousands of Pods, and optimizes the release strategy for The group optimizes resource usage by nearly 25%, saving millions in IT costs every year.
Customer Profile
Jikr Automobile is an intelligent, digital, and data-driven smart travel technology company. Adhering to the concept of user-oriented enterprise, it focuses on the research and development of forward-looking technologies for smart electric travel, and builds a technology ecosystem and user ecosystem. Travel life" is the mission, from product innovation, user experience innovation, to business model innovation, and is committed to bringing users the ultimate travel experience.
In 2021, Jikrypton 001 quickly emerged, setting a record for the delivery of the first model "fastest to break 10,000" in just 110 days. Jikr Automobile, which ranks among the top six sales of new energy vehicle brands in China, launched the Jikr 009 in November last year. It took the lead in completing the first batch of deliveries in just 76 days, setting a new record for the delivery speed of China's luxury pure electric brands. In June of this year, Jikr Auto delivered 10,620 vehicles again, becoming the only luxury pure electric brand that has maintained year-on-year growth for five consecutive months. Jikrypton 001 has become the fastest luxury car in the world to break through 100,000 sales, and once again ranks first in sales of pure electric vehicles above 300,000 yuan.
The company's business is developing rapidly, and the company's IT costs are facing huge challenges. As the technical department that supports the company's many core systems such as contract delivery, vehicle delivery, and payment settlement, the team needs to deal with application releases of different scales almost every day, and the cloud resource consumption required by the application system is increasing day by day.
customer pain point
Multi-cloud hybrid architecture, large number and large-scale K8s clusters, difficult operation and maintenance management
The online and offline use comes from different environments of multiple cloud vendors. With the growth of business in the early stage, the number and scale of container clusters doubled. With a limited team size, it is difficult to take into account business stability and rapid business development. The team faced inconsistent cluster environment configurations, lack of standards for release scripts, and it became increasingly difficult to release and launch applications, which greatly increased the risk of online system stability.
Unpredictable application workloads and lack of data visibility
Kubernetes provides a semantic description of resource requests (Request) for container resource management. Container requests are generally filled in based on manual experience. The administrator will refer to the historical utilization of the container and the pressure test performance of the application, and make continuous adjustments based on the feedback from the online operation.
However, the resource specification configuration mode based on manual experience has the following limitations:
In order to ensure the stability of online applications, administrators usually reserve a considerable number of resource buffers to cope with load fluctuations on upstream and downstream links. The request configuration of the container will be much higher than its actual resource utilization, resulting in cluster resource utilization If it is too low, a lot of resources will be wasted.
When the cluster allocation rate is high, in order to improve the cluster resource utilization, the administrator will actively reduce the Request configuration and coordinate more resource capacity. This operation will increase the deployment density of containers, and will affect the stability of the cluster when the application traffic increases.
The management method that relies entirely on expert experience cannot adapt to the growth of scale, and its management efficiency will decrease as the number of applications increases.
Budget estimation is difficult, and there is a lack of effective means to support enterprise IT fine-grained governance
Under the traditional architecture, the cost amortization of the application is based on the server dimension. However, under the cloud-native architecture, the cost model is no longer completely equivalent to the server. The business team needs more accurate data to estimate the actual input cost of the application, so as to make a reasonable plan for future budget estimates.
solution
Build K8s multi-cluster management in hybrid cloud mode based on ACK One
By registering a cluster with ACK One, users can use ACK's product-based observability in non-ACK K8s clusters, saving the cost of operating and maintaining an external monitoring log system. You can also monitor the log system in a unified way to quickly discover and diagnose problems. At the same time reduce operation and maintenance costs.
View the cost analysis of registered clusters through the ACK console: you can customize the cost of K8s cluster nodes, split costs according to application resource usage, analyze cluster costs, namespace costs, and application costs in multiple dimensions, and quickly locate the source of abnormal costs, helping enterprises Reduce costs and increase efficiency.
Based on ACK FinOps suite insight into resource portraits, improve overall utilization
ACK provides resource profiling capabilities for K8s native workloads. Through the analysis of historical resource usage data, it can provide container-granular resource specification recommendations, which effectively simplifies the complexity of configuring Request and Limit for containers. At the same time, the resource portrait function also provides a visual interactive page on the ACK console, which is convenient for administrators to quickly analyze the rationality of application resource specifications and change resource specification configurations.
Considering that Jikrypton has many online applications, manually adjusting the parameters of different applications brings a heavy workload burden, and the adjusted data cannot be combined with the application release process, which brings hidden dangers to business stability. For this reason, in the process of project landing, ACK resource portrait provides the recommended results of resource specifications in the form of API, and combines with the current release platform of Jikrypton to automatically obtain the optimal value of parameter configuration and render the release template, so that While ensuring business stability, it greatly reduces the overall resource usage cost.
construction results
Based on Alibaba Cloud's cloud-native products, the clusters currently managed by the Jikr technology middle and Taiwan team can not only accurately provide cost data in various business, organization, and resource dimensions, but also greatly improve the cost-effectiveness of using the cloud, improving the company's overall performance by more than 25% The resource utilization is calculated above.