SHENZHEN, China, July 1, 2021 /PRNewswire/ — On the HUAWEI CLOUD TechWave Cloud Infrastructure Day on May 31, Ma Da, Chief Architect of Cloud Container Batch Compute, gave an in-depth introduction to CCE Turbo technologies. CCE Turbo is HUAWEI CLOUD’s revolutionary container cluster and was officially released at the Huawei Developer Conference 2021. CCE Turbo facilitates innovation by accelerating compute, network, and scheduling.
Hardware-software Synergy Accelerates Compute
To optimize the performance of cluster servers, HUAWEI CLOUD released its unique container offload technology based on the hardware-software synergy of the QingTian architecture, and applied the technology to CCE Turbo, fully utilizing cluster resources for service processing. HUAWEI CLOUD also downsized engines, optimized internal logic, and recoded core modules. The shimv2 and agent were recoded based on Rust to reduce processes and memory consumption, improve container startup performance, and enhance Cgroup management capabilities. The high-performance processing capability of the QingTian architecture improves the overall cluster performance by 40% and reduces the resource cost by 30%.
Trunkport Accelerates Network Access
Based on the "IN Cloud" concept of Cloud Native 2.0, HUAWEI CLOUD built a cloud-native network accelerated by the Trunkport technology.
This enables containers to directly access VPCs. The container and VM network layers are flattened into one. Network access time is reduced by half. A thousand container instances can be provisioned within 30 seconds to help customer cope with traffic surges. The network latency is reduced by 40%, making application access smoother.
CCE Turbo is the first in the industry to configure independent security groups and QoS metrics for containers. This better ensures container security and achieves higher forwarding efficiency than other vendors, which puts container and cluster nodes in the same security groups. CCE Turbo supports Kubernetes based on the CRD mechanism to meet complex security isolation requirements and make it easy to configure security groups.
Volcano Accelerates Scheduling
Enterprises need to quickly schedule and fully utilize cluster resources. CCE Turbo provides the following core scheduling capabilities based on Volcano:
- Online and offline hybrid scheduling: CCE Turbo deploys online and offline services in the same cluster, greatly reducing the O&M workload and implementing flexible scheduling. If the access traffic of online services is low, CCE Turbo can use idle resources to process offline compute services (such as analysis and model training). Before the service peak hours, the resources occupied by offline services are automatically released to meet the resource requirements of online services.
- Intelligent scheduling based on application awareness: To further improve the utilization of a hybrid cluster, application awareness models (such as web applications, TensorFlow PS-Worker, and Spark Driver and executor) are used to preempt resources and reuse resources at different time segments based on the resource requirements and application loads of different models. Tasks are sent to the most suitable nodes based on task topologies to reduce latency and data transfer duration, doubling cluster utilization rate.
- Large-scale distributed scheduling: To ensure smooth scheduling of massive tasks in a hybrid cluster, CCE Turbo releases a task scheduler with a distributed architecture to increase throughput. Scheduling algorithms are pruned to narrow down the scope of addressing. The scheduling decision reuse mechanism reduces the scheduling and addressing duration to 1/10 the duration before. Up to 10,000 containers can be scheduled per second.
CCE Turbo Accelerates VIPKID Audio and Video Services
VIPKID has been using CCE Turbo for a long time, and has been benefiting from the acceleration features of CCE Turbo. Ci Yiheng, a senior R&D expert of VIPKID, said in his speech, "CCE Turbo improves our audio and video service performance in every aspect. Our cluster performance is improved by 40%, our service latency reduced by 40%, and our service cost reduced by 43%. CCE Turbo helps us control IT cost while our services are growing fast."
HUAWEI CLOUD is one of the earliest cloud-native vendors and the only Asian founding member of the Cloud Native Computing Foundation (CNCF) community, ranking first in Asia regarding its code contribution and Maintainer seats. HUAWEI CLOUD also contributed the first cloud-native edge compute project KubeEdge, and the first cloud-native batch compute project Volcano. At the Huawei Developer Conference 2021, HUAWEI CLOUD and multiple enterprises jointly released the open-source, cloud-native, multi-cloud container orchestration project Karmada, leading the development of the cloud-native technologies. Regarding product innovation, HUAWEI CLOUD has been the first in the industry to launch a series of cloud-native products and solutions since 2016. In Forrester assessments, HUAWEI CLOUD has won full marks for two consecutive years and ranked first in the container software market in China. To promote industry development, HUAWEI CLOUD and China Academy of Information and Communications Technology (CAICT) jointly released the Cloud Native 2.0 White Paper, which explained the core concepts of Cloud Native 2.0. HUAWEI CLOUD, CNCF, and CAICT established Chuangyuanhui, an international cloud-native communication platform. On this platform, HUAWEI CLOUD will work with cloud-native elites in a range of industries to explore cutting-edge cloud-native technologies, share industry practices, and create infinite possibilities for enterprise digital transformation.
For more information, please visit https://www.huaweicloud.com/intl/en-us/