Recently, Gartner released the “Competitive Landscape: Public Cloud Container Service Market 2021” report. Through this report, Alibaba Cloud became the only Chinese enterprise to be listed for three consecutive years.
According to Gartner, container services have become mature and entered the mainstream market. More diversified container implementation scenarios in enterprises have emerged, and many container vendors have begun mergers and acquisitions. In order to provide container services to more industries, Gartner suggests that vendors should enrich various enterprise-level application image markets. Vendors should also increase support for Application Platform as a Service (aPaaS) and implement various cloud deployment and management. In this report, with the exception of traditional dimensions such as Kubernetes support, container images, Serverless containers, and service mesh, the cluster deployment and management are added for assessment.
Alibaba Cloud is one of the first cloud service vendors in China to provide container services. Alibaba Cloud provides ACK Pro and ACR EE for enterprise-level large-scale production to meet the requirements for reliability and security. Alibaba Cloud also provides services for Serverless containers, machine learning, edge computing, and distributed clouds, which have been applied to enterprises in Internet, new retail, education, healthcare, and other industries.
Serverless Containers with Simple management and Advance Scaling
Serverless technology has become a hot topic in recent years. The main aim of serverless is to help developers reduce workload by minimizing routine code development tasks, such as for server O&M, for cloud applications. Through serverless, developers can focus more on creating more value for businesses. At the same time, serverless technology can also improve the startup speed of applications significantly.
Apart from Function Compute (FC) for event-driven applications and Serverless Application Engine (SAE) for microservice applications, Alibaba Cloud also provides Serverless container products called Alibaba Cloud Serverless Kubernetes (ASK). ASK provides a standard Kubernetes interface and extreme elasticity for users without O&M requirements. Users can conveniently create Serverless clusters without the need to manage Kubernetes nodes and servers. Users do not need to carefully plan the capacity or care about the underlying servers.
Pods in ASK clusters run in secure and isolated container environments based on Elastic Container Instance (ECI). The underlying compute resources of different ECIs run independently as they are completely isolated by lightweight virtual sandboxes. In addition, based on the Serverless container architecture deeply optimized by Alibaba Cloud, ASK clusters are provided with great elasticity without being limited by the computing capacity of cluster nodes. Currently, ASK clusters can start 500 pods in 30 seconds, and a single cluster can provide a capacity of 20 KB for each pod.
Prudencemed, a mobile medical device developer, uses Serverless containers to improve the prediction accuracy of AI models to 99.99%. Weibo uses Serverless containers to implement auto scaling of business, reducing the end-to-end scaling time by 70%.
Two Cutting-edge Scenarios: Machine Learning (Cloud Native AI) and Edge Computing
Tailored to provide more optimization space for AI engineering efficiency based on cloud native
Alibaba Cloud container services have been implemented in cloud native AI scenarios for the past few years. Alibaba Cloud has provided many tools, services, and solutions for heterogeneous resource management, AI application lifecycle management, AI task scheduling and elasticity, AI data management and acceleration, and unified AI-big data orchestration. It has helped many customers improve the production efficiency of AI engineering, optimize GPU resource utilization, and accelerate AI platform construction. In addition, Alibaba Cloud promotes the development of the Kubernetes community towards large-scale data computing fields, such as AI and big data.
TAL, an online education enterprise, deploys a large number of important AI+ education businesses through Alibaba Cloud container services. By doing so, TAL has accumulated over 100 AI capabilities and over 10 AI solutions for education scenarios in terms of images, audios, and natural language processing, covering all teaching stages. With these capabilities and solutions, TAL has improved the overall AI development and deployment efficiency and realized efficient MLOps. In addition, TAL has solved the problem of low resource reuse rate in AI service scenarios requiring high concurrency, high performance, and high resources. The stability of the platform reached as high as 99.99%.
At the Alibaba Cloud Computing Summit 2021 in Shenzhen, Alibaba Cloud announced the cloud-native AI solution. This solution provides several features such as underlying resource O&M and optimization, multi-tenant quota management, AI task management and intelligent scheduling, big data service integration, and support for AI engine acceleration and algorithm application at upper layers. At the same time, it provides a visual interface, command line tools, and SDKs, which are simple and easy for extension and integration. AI service providers, including data scientists, AI algorithm engineers, and AI platform builders and operators, can freely use these features to customize their own AI platforms based on Kubernetes.
Edge cloud native runs Kubernetes outside the cloud
Traditional cloud computing centers adopt a centralized storage and computing model, which cannot meet the needs of edge devices for timeliness, capacity, and computing power. Therefore, building edge infrastructures has become a new topic.
The cloud-native architecture of cloud-edge-device collaboration has been developed. On the cloud, the cloud-native management and product capabilities are retained. A large number of edge nodes and edge businesses have become the workloads of the cloud-native system. To achieve the integration of the business, O&M, and ecosystem, edge cloud native technology provides better service governance and traffic control capabilities, unified application management and O&M, and higher isolation, security and efficiency.
In 2019, Alibaba Cloud released ACK@Edge, the first container product for edge computing in the industry. ACK@Edge features cloud standard management and edge moderate autonomy. On the cloud, it offers powerful cloud-native management capabilities, which can achieve business integration such as CDNPaaS and IoTPaaS. Cloud-edge O&M and management can also be implemented through multi-specification and multi-link solutions. On the edge, based on native Kubernetes capabilities, ACK@Edge provides capabilities such as edge autonomy, unitized management, traffic topology, and fine-grained detection of edge computing power.
SUNAC group builds a digital and intelligent customer service platform for community customers based on ACK@Edge. On this platform, SUNAC uniformly manages intelligent devices in community scenarios such as gateway systems and information screens, and in home scenarios such as TV, smart air conditioners, and Tmall Genie. For example, SUNAC achieves the all-in-one management of community parking lots and enables intelligent collaboration based on cloud-edge-device unification. Youku uses ACK@Edge to uniformly manage a dozen regions and numerous edge nodes on the public cloud. Through elastic scaling, Youku saves machine costs by more than 50% and reduces device-to-device network latency by 75%.
Distributed cloud breaks the boundary to achieve unified and fine-grained management
After the significant changes in the competitive landscape for container management in 2020, mainstream cloud service providers have released container management solutions of distributed clouds. Gartner predicts that 81% of enterprises will adopt the multiple/hybrid cloud strategy. “Most enterprises adopt the multi-cloud strategy to avoid supplier locking or using the best solution.” This strategy ensures the purchase agility of enterprises, and balances availability, performance, data sovereignty, regulatory, and labor costs.
Alibaba Cloud has provided hybrid cloud support for ACK since September 2019. In September 2020, ACK upgraded the management capability of distributed cloud applications. In addition to Kubernetes clusters on Alibaba Cloud, users can also manage Kubernetes clusters in their own IDC and on other clouds. Thus, centralized cluster management, IT governance, security protection, application management, and backup management can be realized. This year, Alibaba Cloud also released unified delivery capability of application centers for ACK. Developers can use GitOps to publish applications to different cloud environments in a secure, consistent, and stable manner. This ensures that the whole process is secure and controllable, and improves the efficiency and stability of application delivery.
ACK can be used in combination with Alibaba Cloud Service Mesh (ASM). As the first Istio-compatible hosted service mesh platform in the industry, ASM provides unified capabilities for services running on different computing infrastructures. Through traffic control, mesh observation, and inter-service communication security, ASM enables features such as service nearby access, failover, and gray release. ASM can simplify service management and be applied to Kubernetes clusters, Serverless Kubernetes clusters, Elastic Compute Service (ECS) instances, and clusters created by users.
ASM insists on developing “hosted, standard, secure, stable, simple, and scalable” products. Currently, it has covered 12 regions and maintains the leading position in product integrity and market share. ASM has been used in the largest mesh cluster in the industry and core businesses such as e-commerce and DingTalk of Alibaba. In addition, it also continues to rank first in terms of number of customers for public cloud, covering online education and vocational training, e-commerce, logistics, supply chain, gaming, Internet of vehicles, and IoT.
ASM decouples service governance from business and submerges to the infrastructure layer, so that business development focuses more on the business. By doing so, in a standardized way, ASM solves the challenges of multi-language multi-programming framework and service governance in the microservice software architecture. In ultra-large scale application scenarios, ASM ensures system stability, improves observability and diagnosability, and improves service governance capabilities and performance.