This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search


Is the Diskless Architecture the Most Preferred for the Cloud and Internet Industries?


The explosive growth of mass data poses challenges to cloud and Internet data centers. However, the rapidly emerging diskless architecture helps data centers achieve energy saving, high performance, and high reliability in the multi-cloud era.

Challenges for Traditional Cloud and Internet IT Architectures

In China, the cloud and Internet industries have created the largest IT infrastructure platform, which stores and processes a greater amount of data than any other industry. As the scale of data center construction expands, it becomes crucial to develop efficient approaches to handle the surge in construction demands, adhere to green and energy-efficient requirements, and ensure the high-quality evolution of data centers.

The traditional data center typically adopts a multi-level hierarchical architecture. Each layer from the server, network, to the storage layers is a complete computer system that consists of components such as CPU, memory, bus, and hard disks. In response to new data applications, organizations often adopt an integrated server architecture that couples applications with local disks to enable the rapid deployment and testing of new services. However, this conventional server (local disk) architecture can result in an imbalanced allocation of hardware resources, with computing and storage resources developing at different rates. The computing power life cycle and data life cycle are increasingly different. This mismatch can lead to issues such as rigid expansion, idle resources, and low utilization, which hinder service data sharing and access requirements.

Rapid Emergence of the Diskless Architecture

The data center industry is facing obstacles in optimizing capacity utilization and storage efficiency. However, the development of new hardware technologies, such as high-performance disk enclosures, dedicated data processors (such as DPU), remote direct memory access (RDMA), and CXL protocols, provides a technical basis for reconstructing data center infrastructure. By adopting these innovations, new data centers can fulfill the demands for resource utilization, reliability, performance, and efficiency.

The diskless architecture is quickly gaining popularity due to the emergence of new service challenges and the rapid advancements in hardware technology. This architecture involves deploying local disks of servers separately to allow the decoupling and reorganization of multi-level hierarchical resources, enabling independent expansion and flexible sharing of different types of hardware. Moreover, it offloads tasks that general-purpose CPUs cannot perform to dedicated data processors, thereby breaking away from the conventional CPU-centric processing logic. Therefore, enterprises can optimize the usage of hardware resources on a global level and achieve optimal energy efficiency ratios in their hardware combinations.

OceanDisk Smart Disk Enclosure
Building a Storage Base for Data Centers in the Multi-Cloud Era

In response to the growing popularity of diskless server architecture, Huawei has launched the OceanDisk Smart Disk Enclosure. The product helps cloud and Internet data centers achieve green and energy-efficient management, high performance, and high reliability in the following ways:

• OceanDisk connects to servers through NoF high-speed Ethernet networks and offloads complex data storage capabilities to smart disk enclosures. This approach enables OceanDisk to deliver performance levels comparable to that of local disks.

• The diskless server architecture fully utilizes the advantages of decoupled storage and compute. It implements independent storage and compute elasticity, and does not require data migration during CPU computing power upgrade.

• The large-ratio erasure code (EC) and deduplication and compression capabilities reduce cabinet space and power consumption each by 40%.

• Multiple reliability technologies, such as disk sub-health management and intelligent slow disk optimization, are used to achieve system-level reliability. Data disk faults can be predicted, visualized, and managed, greatly simplifying O&M.

The challenges of rapid data growth and the need to reduce costs and improve efficiency will persist alongside new services like cloud and internet. The unstoppable trend towards an efficient, intensive, and decoupling diskless architecture for data centers presents new opportunities for industry transformation. This diskless architecture is flexible and allows for refined resource utilization, as well as green and low-carbon energy consumption. This will help data centers reduce computing power waste and achieve green, intensive development, opening up new possibilities for industry transformation.

Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy, position, products, and technologies of Huawei Technologies Co., Ltd. If you need to learn more about the products and technologies of Huawei Technologies Co., Ltd., please visit our website at or contact us.