Traditional server-based datacenters suffer from low resource utilization, insufficient hardware scalability, insufficient resource usage elasticity, and low fault tolerance. With the ever-increasing network speed and processing power in hardware controllers, physically disaggregating the coupled resources in a server has been a promising trend for building next-generation Datacenters, as it effectively addresses the limitations of server-based datacenters. Although some existing works follow this trend, how to build an efficient physically resource-disaggregated data center (PRD-DC) is still in the exploring stage.
Thus, a research team led by Haiqiao WU published a comprehensive review on 15 May 2026 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.
The team firstly describe the trends and technique routes for PRD-DCs. From the perspective of resource management, there are mainly 3 technique routes to realize a PRD-DC, including CPU-centric, Memory-centric, and Resource decentralization. Under the CPU-centric technique route, all management and usage logic of storage and computing resources run on the CPU, and other disaggregated resources are used by accessing remote resources. The memory-centric technique route migrates the memory management functions in traditional OS to the memory side. However, the resource management logic of the above two technique routes is centered on a certain resource, where other resources are still subject to certain types of resources. Thus, resource decentralization emerges due to its high scalability and elasticity of all resources, which decouples the traditional OS completely and runs the decoupled management functions of different resources on the corresponding resource side.
Further, the team proposes a potential solution for realizing a PRD-DC, named Serverless DC, which follows the resource decentralization technique route. In Serverless DC, all kinds of storage/computing units are connected to the low-latency and high-throughput network through IPU (I/O Processing Unit), and managed by SDC-kernel. The units under an IPU form a physical resource pool. IPU is responsible for connecting, operating, and scheduling the disaggregated storage/computing resources and provides a unified network for the interaction between the resources, including a control plane, data plane, and IO subsystem. The SDC-kernel is an OS architecture designed for disaggregated resource management in Serverless DC, and owns the following features:1) stateless kernel functions decoupling, 2) hierarchical resource management, and 3) diverse resource use interfaces, which are realized through the following designs, including Memory positioning, Control Center Module, Resource Management Module, and Interruption management. Due to the longer communication distance in PRD-DCs, Serverless DC reduce the latency from the following aspects, involving transmission medium, switch strategy, interconnection topology, out-of-band control, and network stack.
Finally, the team concludes several open research issues in the field and presents some potential solutions, including compatibility for applications, reliability and security, power supply and energy-saving, heterogeneity supporting.
DOI:10.1007/s11704-025-41118-5