Connect with us


Huawei openEuler 21.03 brings new 5.10 kernel and supports more desktop environments



At Huawei Developer Conference 2020 (Cloud), the openEuler community has released its first long-term support (LTS) edition, the openEuler 20.03 LTS. This announcement marks a new milestone for the commercial readiness of openEuler-based operating systems.

openEuler is the open-source free Linux distribution platform that offers an open community for developers worldwide and set to take the multi-core heterogeneous computing industry to the next level.

At the same time, Huawei has joined with four leading OS vendors, Kylinsoft, iSoft, Uniontech, and the Institute of Software Chinese Academy of Sciences (ISCAS), to announce commercial editions powered by openEuler.

Now, according to the latest news, the latest edition of openEular has arrived, openEular 21.03, which is more innovative than its processor LTS version. It uses a  brand new 5.10 kernel, enhanced virtualization capabilities, and maintenance capabilities, and also brings more desktop environment support.

OpenEuler 21.03 update details:

Brand new 5.10 kernel:

Deeply optimize scheduling, IO, memory management, and provide more computing power support such as Arm64, x86, and RISC-V.

  • The scheduler is optimized to optimize the fairness of the CFS Task, and the numa aware asynchronous call mechanism is added, which has a significant improvement in NVDIMM initialization; optimizes the scheduling strategy of SCHED_IDLE, significantly improves the scheduling delay of high-priority tasks, and reduces the cost to other tasks interference.
  • The optimization of numa balancing mechanism brings better affinity, higher utilization rate, and fewer invalid migrations.
  • The CPU isolation mechanism is enhanced, interrupt isolation is supported, unbound threads isolation is supported, and the isolation of CPU cores is enhanced, which can better avoid mutual interference between services.
  • The performance of cgroup single-thread migration is optimized to eliminate the dependence on the thread group’s read and write semaphore; the introduction of Time Namespaces makes container migration more convenient.
  • The system container supports the ability to limit the number of file handles used in the container. File handles include ordinary file handles and network sockets. When starting the container, you can specify the –files-limit parameter to limit the maximum number of open handles in the container.
  • PSI capability support, PSI (Pressure Stall Information) provides a method to evaluate the pressure of system resources such as CPU, Memory, and IO. Accurate detection methods can help resource users determine the appropriate workload, and can also help the system formulate efficient resource scheduling strategies, maximize the use of system resources, and maximize user experience.
  • Inter-process communication optimization, pipe/epoll_wait wake-up mechanism optimization, solve the performance problem of waking up multiple waiting threads.
  • Enhanced memory management, refined memory control, statistics, heterogeneous memory, hot-swap, memory initialization, and other functions have been improved, and provide a more effective user control interface; hotspot lock and semaphore optimization, radical memory regulation and defragmentation Optimize the vmap/vmalloc mechanism to significantly improve the efficiency of memory application; KASAN, kmemleak, slub_debug, oom, and other memory maintenance features have been enhanced to improve memory problem location and resolution efficiency.
  • The Early Departure Time model switch solves the limitation of the original TCP framework in the packet sending process. According to the scheduling strategy, the EDT timestamp is set for the data packet to avoid the delay caused by the large queue buffer and bring about a large improvement in TCP performance.
  • MultiPath TCP support can improve performance and reliability in mobile and data scenarios, and supports parallel transmission of multiple substreams in load balancing scenarios.
  • The log fast commit method is introduced, and EXT4 introduces a new and more lightweight logging method-fast commit, which can greatly speed up time-consuming operations such as fsync and bring better performance.
  • Support dm write-cache feature, improve SSD bulk sequential write performance, and improve the performance of DDR persistent memory.
  • Support io_uring, io_uring is a new asynchronous IO framework and implementation that supports polling mode. In polling mode, the performance is improved significantly, which is close to spdk, and the performance is better when the queue depth is high.
  • Support ILP32, support 32-bit applications on Kunpeng 920 Arm64 environment.
  • IMA commercial enhancement, based on the open-source IMA solution, enhances security, improves performance, improves ease of use, and facilitates commercial implementation.
  • Support per task stack inspection to enhance the protection against ROP attacks.
  • MPAM resource management and control, support Arm64 architecture Cache QoS and memory bandwidth control technology.
  • Supports SEDI-based NMI mechanism and PMU-based NMI mechanism, enabling hard lockup detection; enabling perf nmi enables more accurate performance analysis.
  • The Arm64 platform supports virtual machine CPU hot-swap, which improves the flexibility of resource configuration.
  • Arm64 kdump is enhanced to support memory reservation for addresses above 4G, so that kdump can reserve more memory space and support machines with larger memory.
  • It supports Raspberry PI series boards. The support of Raspberry Pi has been integrated into the native openEuler 21.03 kernel, and you can directly use the openEuler 21.03 kernel source code for debugging.
  • RISC-V platform supports KVM virtualization
  • Support 1822 smart network card

Kernel Hot Upgrade

Kernel hot upgrade is an OS bug fix and upgrades solution, which realizes fast hot replacement of the kernel without business perception.

  • The features of Cpu Park and Quick Kexec speed up system startup and shutdown, reduce downtime, and increase system availability.
  • Pin Memory and Pmem features ensure rapid and accurate recovery of business processes and improve business resilience.
  • The kernel hot upgrade controller provides a gRPC communication interface, which is easy to use.

Memory Grading Expansion:

Supports unified management of multiple memories and storage media, and smooth expansion of system capacity.

  • Hot and cold page identification, through the kernel state memory page free and idle statistics mechanism, accurately identifies the hot and cold distribution of process memory page access.
  • The elimination strategy can be configured, the configuration interface is provided, and the memory page hot and cold classification strategy can be customized.
  • Smooth expansion, cold pages are automatically swapped out to expanded memory, and the software deployed on it runs compatible without changing and adapting the programming mode.
  • Multi-media expansion support, support SCM, XL Flash, NVMe SSD, and other media as expansion memory, specify the memory hot and cold layering scheme according to the access speed of the media itself, to achieve the purpose of expanding memory and reducing performance loss.

Enhanced virtualization capabilities and maintainable measurement capabilities:

Increase the capability expansion of Hot Migration Pro to improve the maintainable and measurable capability. Hot migration Pro features, enhanced hot migration multifd support TLS, to ensure data security during the migration process; support for parallel compression of hot migration data to improve migration performance; increase statistics on the frequency of data page access, and support advance prediction of hot migration data.

The performance debugging tool (vmtop) can dynamically view the resource usage of the virtual machine in real-time, including CPU usage, memory usage, and other information. Added extension to support x86_64 architecture. IO suspension support, IO will automatically retry by default when an IO error occurs, and an alarm will be reported on timeout. RISC-V architecture supports virtualized hot migration

Lightweight virtual runtime (Stratovirt):

Increase elastic memory, large page function, system call filtering function, and enhance the IO subsystem to improve performance and stability.

  • Flexible memory support, according to the memory requirements of the workload, realizes the allocation and recovery of memory, and the virtio-balloon memory recovery speed is up to 3GB/sec.
  • Large page support. Large page support is provided under the lightweight framework, which can provide continuous physical memory pages for lightweight virtual machines and improve the memory access efficiency of virtual machines.
  • System call filtering, simplifying the device model, adding system call filtering support, only 35 system calls are needed in the simplest configuration, effectively reducing the system attack surface.
  • The IO subsystem is enhanced to support multi-channel concurrent IO capability to improve performance; to support IO-QOS capability to enhance the flexibility and stability of virtual machine IO traffic management.

OpenStack Victoria integration:

Simple, scalable, rich, and unified cloud management operating system. For more features, please refer to the official release notes of OpenStack Victoria.

  • Integrate the OpenStack Victoria version to enable infrastructure as a service (IaaS) solutions.
  • Enhance block storage service capabilities and add advanced functions such as capacity expansion, snapshots, and virtual machine image cloning.
  • Enhance containerized deployment and network capabilities to better integrate with containers.
  • Added extended service support, supporting extended services such as control panel management, bare metal deployment, and cloud resource tracking.

Kubernetes 1.20 integration:

For more features of the cloud-native operating system for automatic deployment, expansion, and management of containerized applications, please refer to the official release notes of Kubernetes 1.20.

  • Automatic online and rollback. Kubernetes will automatically bring the application or its configuration changed instance online, and monitor the running status of the application. If it fails, the previous changes will be rolled back.
  • Service discovery and load balancing, service discovery, and load balancing machine support based on container IP and DNS name.
  • Storage orchestration supports the automatic mounting of multiple storage backends, such as local storage, NFS, iSCSI, Gluster, Ceph, and other network storage systems.
  • Horizontal expansion, support command line, UI manual operation expansion, and automatic expansion based on CPU usage.

HA high-availability cluster solution:

The high-availability cluster solution contributed by Kylin Software, with second-level switching of failures, provides users with a highly available environment with a business continuity guarantee, continuous data protection, and disaster recovery.

  • It supports multiple protection modes, such as dual-system hot backup, dual-system mutual backup, multi-system backup (N+M), and other protection modes to meet various protection needs of business applications.
  • Both physical machine scenes and cloud scenes are supported, and high availability can be configured in the virtual machine pool, and high availability protection of physical server nodes can also be carried out.
  • Support network (including single heartbeat and dual heartbeat modes) and other heartbeat modes, comprehensively monitor shared data resources and ensure data consistency in extreme cases.
  • The resource consumption is low, the software consumes fewer system resources, and has little impact on application deployment and resource usage security.
  • Support mainstream system services and application software, such as nginx, httpd, mariaDB, etc., and support secondary development.
  • Support HA-WEB human-computer interaction interface, such as user login, cluster status display, resource control, etc.
  • Support HA-API cluster control REST interface, such as back-end cluster control, resource management, cluster status monitoring, resource status monitoring, etc.

More desktop environment support:

Provide more development desktop choices and a better development experience.

  • Xfce desktop support. Xfce is a lightweight Linux desktop that occupies fewer resources than mainstream UI.
  • DDE desktop support is a Linux desktop environment owned by Tongxin Software, which is beautiful, easy to use, and efficient.

[Via – ithome]

Most of Deng Li's smartphones are from the Huawei ecosystem and his first Huawei phone was Ascend Mate 2 (4G). As a tech enthusiast, he keeps exploring new technologies and inspects them closely. Apart from the technology world, he takes care of his garden.