–
Moscone Center
The rise of Artificial Intelligence (AI) has sparked a wave of innovations across the entire optical industry. The demand for efficient optical I/O to connect compute clusters in data centers is driving rapid technological advancements. AI has become an indispensable tool for automating and optimizing network management, potentially reducing operational costs significantly.
To fully embrace AI and address challenges, networks must evolve to efficiently handle joint scheduling of traffic across server communications, data center interconnects, edge computing, and customer data flow. Although hyperscalers and network operators differ in topologies, they face similar challenges in meeting the demands. In fact, modern data center design calls for dynamic topology optimization together with agile routing. A new frontier is the joint scheduling and reconfiguration of data center networks in conjunction with computing to optimize the time delay. This complicated tandem optimization will challenge analysis and is ideal for AI assist.
Although incorporating AI can enhance network efficiency and reduce operational costs, implementing and maintaining AI algorithms come with its own cost. The required computational resources could lead to increased energy consumption. AI has created new attack surfaces, so safe-AI must be practiced to ensure secure network operations.
This summit will discuss the following topics in future network architecture design for hyperscale operators and network operators:
- Key requirements and challenges;
- Design and optimization to leverage AI while minimizing its environmental impact;
- Agile network topology to manage power consumption hot spots and cold spots;
- Trade-offs between meeting minimum requirements and achieving acceptable performance gains;
- Necessary changes and improvements in network equipment; and
- Vulnerabilities exposed by AI, guard-rails or smarter monitoring to mitigate attacks.