Key Features

Tiered Memory Expansion

Today's memory manufacturing processes have nearly reached the peak of how advanced they are. The mature ARM ecosystem is making the cost per CPU core lower and lower. Databases, virtual machines (VMs), big data, artificial intelligence (AI), and deep learning call for increasingly higher computing power and memory capacities. Limited memory capacities have become a challenge for service growth. Tiered memory expansion has proven itself as a solution for that challenge. DRAM and low-speed memory media, such as storage class memory (SCM), Apache Pass (AEP), and remote direct memory access (RDMA), form a multi-tier memory structure. Automatic memory scheduling redirects hot data to the DRAM high-speed memory area and cold data to the low-speed memory area. The tiered memory structure increases the memory capacity and ensures efficient and stable running of core services. This feature is ideal for applications that use a large number of memory cache and access models randomly. Tests show that it delivers 40% higher performance for MySQL than counterparts. The user-mode memory swapping mechanism is added for the user-mode storage framework and user requirements.

• Process-level control: Supports processes that use configuration files to expand the memory. Compared with the native LRU-based pageout kswap mechanism of Linux, it is more flexible and accurate.
• Cold and hot tiering: In user mode, a memory access scan can be performed for a designated process. You can select a cold and hot tiering policy configuration file to classify obtained memory access results into hot memory and cold memory.
• Discarding policy: The cold memory is discarded when it meets the conditions specified in the etMem configuration file and the system environment configuration. The discarding process uses the native kernel capability, which is secure and reliable and does not affect user experience.
• Multi-media expansion support: Multiple media, such as SCM, XL flash, and NVMe SSDs, can be used as the expanded memory. The cold or hot memory tiering solution is specified based on the access speed of the medium to expand the memory while reducing the performance loss.

Application scenario: tiered memory expansion for service processes on a node. Ideal for applications that use a large amount of memory but do not access the memory frequently, such as MySQL, Redis, and Nginx. All memory expansion operations are performed within a node and no cross-node operations are involved. In a user-mode storage framework, the userswap function can enable user-mode storage as a swap device.

KubeOS Container Operating System

Cloud native is the next step in the evolution of cloud computing. Kubernetes has become the foundation of cloud native software infrastructure. Major OS vendors have launched their OSs for cloud native scenarios, such as Red Hat Enterprise Linux CoreOS (RHCOS) and AWS Bottlerocket. These new OSs are deployed and managed in containers, delivering an O&M experience similar to service containers. To adapt to the cloud native trend, openEuler launches KubeOS, a container OS that centrally manages cloud native cluster OSs in containers. KubeOS has the following features:

• OS containerization and Kubernetes interconnection for atomized lifecycle management
• Lightweight OS tailoring to reduce unnecessary packages for quicker upgrades and replacements

Application Scenario: OS Management of containerized cloud service hosts based on Kubernetes, delivering a lifecycle management and O&M experience similar to container services.

eggo Kubernetes Deployment Tool

eggo is a Kubernetes cluster deployment and management project of the openEuler cloud native special interest group (SIG). It provides efficient and stable cluster deployment capabilities. eggo supports multiple architectures for a single cluster as well as online and offline deployments. It incorporates GitOps to detect cluster configuration changes and enable unified and efficient deployment of cluster OSs.

• Version-based cluster configuration management: Git repositories are used to store and track cluster configuration changes.
• Configuration awareness: GitOps detects cluster configuration changes in Git repositories, and then sends a cluster operation request to the deployment engine.
• Deployment engine: eggo delivers tasks to the service cluster. For example, it triggers tasks such as deploying the service cluster, destroying the service cluster, adding nodes, and deleting nodes.

Application Scenario: x86 and ARM dual-plane cloud infrastructure. Based on the native Kubernetes cloud native framework, eggo implements unified cluster deployment, monitoring, and audit of OSs.

A-Ops

A-Ops is an OS-oriented O&M platform that provides intelligent O&M solutions covering data collection, health check, fault diagnosis, and fault rectification. The A-Ops project includes the following sub-projects: fault detection (Gala), fault locating (X-diagnosis), and defect rectification (Apollo).

• Community hot patch pipeline

  Hot patch preparation: The target version and patch file of the software package can be specified in the cold patch pull request (PR) to make a hot patch.

  Hot patch release: Hot patches to be released can be automatically collected based on hot patch issues, and then released by reusing the cold patch update release logic.

• Enhanced vulnerability management

  Intelligent patch inspection: CVE inspection and notification for a single-node system or cluster, and one-click fix and rollback are supported.

  Hot fixing: Some CVEs can be fixed using hot patches, ensuring zero service interruption.

  Patch service: Cold and hot patch subscription allows patches to be acquired online.

• da-tool

  da-tool is a latency analysis tool based on kprobe. It streamlines the configuration process of kprobe and displays the latency characteristics of traced functions. da-tool suppotrs:

  • Analysis of latency characteristics of UDP receiving and sending functions

  • Analysis of UDP-related process scheduling

OpenStack Train Support

OpenStack Train is a simple, scalable, rich, and standard cloud management operating system. For details about more features, see OpenStack Train release notes.

• OpenStack Queens/Rocky Integration, which enables IaaS solutions.
• Enhanced block storage. Advanced functions such as capacity expansion, snapshots, and VM image cloning are supported.
• Container-based deployment and network capabilities. Better integration with containers is achieved.
• Extended services. Extended services such as control panel management, bare metal server deployment, and cloud resource tracing are supported.
• OpenStack Train. Neutron port forwarding and Ironic DPU offload are added. For details about the patch, see https://atomgit.com/openeuler/openstack-plugin/tree/train/.

Desktop Environment Support

• Kiran desktop environment, developed by Kylinsec, is a stable, efficient, and easy-to-use desktop environment oriented towards user and market requirements. It consists of the desktop, taskbar, tray, control center, and window management components.

Compatibility List

• Northbound Compatibility List
• Intel Ice Lake supported.

Automatic Optimization for Software Package Downloads

30 openEuler mirror sites are distributed across Asia, Europe, and North America. Software packages can be downloaded from the nearest mirror sites to improve the download speed.

A metalink, whose value is the URL of the API provided by the metalink service, is configured in the DNF or Yum configuration file shipped with openEuler releases. When a user tries to download a software package, the DNF or Yum client sends a request to the metalink URL. The metalink service returns data in XML format that contains the addresses of nearest mirror sites. The DNF or Yum client then selects the optimal site from the addresses to download the software package, ensuring a fast download speed.