Challenge 1: Data Infrastructure with Maximum Energy Efficiency

To handle various devices and diverse application workloads in data centers, high-throughput data processing needs, and energy saving trends, research is needed into technologies such as data-centric, network-storage-compute converged architecture, diversified heterogeneous virtualization, data center-level energy saving, data-storage collaboration, and data services that deliver optimal user experience. This will enable us to build a data infrastructure with optimal energy efficiency.

Challenge Direction 1: Data-Centric, Network-Storage-Computing Convergence

1. Storage-compute offloading and acceleration based on DPUs: Explore methods for offloading storage services to DPUs to accelerate data access, and develop DPU solutions to accelerate mainstream applications like big data and AI.

2. Storage-network offloading based on network devices: Explore methods for converging network and storage address spaces, high-performance distributed caching, redundant distribution of stored data, as well as data stream processing mechanisms and models.

3. Point-to-point, high-throughput data transmission: Explore pass-through protocols and link control between storage media and optical networks to solve the bottleneck of point-to-point, long-distance big data transmission.

Challenge Direction 2: Key Technologies for Multi-Cloud IT Architecture

1. Multi-cloud container and storage management & scheduling: Research a unified scheduling algorithm for optimal container deployment and storage resource allocation across domains in multi-cloud interconnection scenarios that involve public clouds, private clouds, and traditional data centers, thereby enabling applications to share on-premises and cloud infrastructure resources.

2. Multi-cloud data sharing and mobility: Research metadata-based global indexing to implement efficient data sharing and mobility across domains and clouds in scenarios that involve interconnection between devices, edges, and public & private clouds, thereby helping customers extract value from both on-premises and cloud data.

3. Lightweight cloud-native virtualization: With the focus on serverless cloud-native technologies, research ultra-lightweight resource virtualization and isolation technologies that are compatible with existing virtual machine and container technologies while delivering good isolation and security, thereby enabling a customer's device, edge, and cloud infrastructures to adopt a unified architecture with secure and efficient resource utilization.

Challenge Direction 3: Energy-Saving Technologies for Data Infrastructure

1. Hardware-level energy saving: Explore efficient heat dissipation technologies (such as mixed-flow fan) oriented to high-density and high-impedance storage hardware, develop reliable, energy-efficient, non-water liquid-cooling working media and optimal deployment solutions with full liquid cooling of storage nodes, and research energy-saving technologies such as waste heat recovery, new-energy power supply, high temperature resistance, unattended operation, and high-performance storage media with nearly zero power consumption in idle state, thereby reducing the hardware power consumption of a storage system in an end-to-end manner.

2. Energy saving via software-hardware collaboration: Explore working models that help storage systems save energy, including but not limited to adjusting the running and hibernation states of internal components based on changes in the amount of workloads to minimize energy consumption in non-working hours. Develop an efficient architecture that uses heterogeneous computing collaboration to minimize processor energy consumption and compilation-based energy consumption control to minimize the energy consumed by program running.

3. Optimal energy efficiency throughout the data lifecycle: Explore data storage and access models that maximize energy efficiency throughout the data lifecycle, including but not limited to cross-site or multi-cloud data tiering and energy-efficient data prefetching. Explore the optimal HDD-based model for cold data distribution and scheduling, and balance the read/write performance, the number of powered-on devices, and energy consumption, thereby achieving the maximum energy efficiency of data capacity under ultra-low power consumption. Use optical-electrical hybrid simplified networks and energy-efficient data transmission protocols to achieve the maximum energy efficiency of data transmission.

Challenge Direction 4: Digital Resilience and Proactive Protection Technologies

1. Storage protection against malicious attacks from viruses like ransomware: Research technologies such as real-time attack detection (like ransomware attacks), slow attack discovery, damage analysis and locating, content association, and I/O-level snapshot recovery to build capabilities like response to attacks in subseconds, zero data loss, and full recovery.

2. Native secure storage: Research proactive identification of storage system risks, quick detection of threats and attacks, hardware-level offloading of security algorithms, hardware-based control over critical data access, AI-powered user anomaly detection, response and recovery, and online self-training of AI-powered detection models to ensure no impact on resources and services and build native data security capabilities.

3. Security in multi-cloud scenarios: Research key technologies such as data and metadata encryption, sensitive data identification, data asset discovery, cross-domain data mobility control, privacy computing, and data capsule in multi-cloud scenarios to ensure secure and trustworthy cross-domain data mobility.