Cisco highlights three directions in network development that will dominate the next twelve months. The author of the forecast is Jonathan Davidson, Executive Vice President and General Manager of Cisco Networking.
Forecast 1: Evolution of Ethernet. There will be ongoing innovations that reimagine Ethernet technology to deliver greater efficiency and resilience as it essentially becomes the building block for AI-powered networks.
Most customers want to expand their networks to support new types of workloads and their increasing numbers. At the same time, they expect a single architecture that reduces operational complexity and costs – a single network infrastructure to handle everything. Ethernet has proved to be the preferred technology for training artificial intelligence and improving AI reasoning, providing a standardized, flexible, and affordable alternative to proprietary technologies such as InfiniBand.
To understand why, it is necessary to recall the evolution of Ethernet and the role that technology has played in pioneering other important technological changes. Traditional loads such as audio, video, data storage, high-performance computing, and high-performance databases have required Ethernet to become ‘lossless’ and significantly improve reliability and performance while reducing latency. Ethernet also makes economical scaling possible to large central clusters and smaller edge clusters.
The importance of versatile switching solutions will be evidenced in the evolution of corporate networks supported by artificial intelligence towards simultaneously handling traditional loads and those associated with AI and machine learning. New ideas for solving I/O operation related problems will result in cycles of proprietary technology development and standards, resulting in changes to the landscape of internal server and inter-server structures. For example, faster and more efficient data transfer between models and data warehouses will assimilate graphics processors to basic mass memory systems, and server design will blur the boundaries between internal and external structures.
Forecast 2: The modernization of network infrastructure and data centers will remain an investment priority for companies seeking to keep up with the increase in loads, boost performance, and operational efficiency.
The 100G network technology is currently widespread, but it’s not difficult to predict that the industry will soon reach a critical point in which network architectures will require more performance in order to support new applications, data, and workflows, including artificial intelligence and machine learning. The demand for higher network throughput is increasing, driven by innovations in graphics processors and network cards, whose speeds are already reaching 400G and 800G.
This need represents a fundamental change in how data centers are built and function. The development of 400G/800G “leaf-spine” architectures with 100G SerDes arrangements responds to increased demand for greater server-side and client-side throughput. We will also observe growing pressure for security and multi-access to segregate data in terms of AI workloads. Moreover, customers will apply liquid cooling systems to ensure optimal energy consumption by AI infrastructure where possible.
Environmental, Social, and Corporate Governance (ESG) initiatives are the companies’ top priorities currently, which has created additional pressure on IT organizations, including networking teams, to support sustainable IT and data center initiatives. Modernizing network infrastructure and data centers can be a key step, as many of the latest products have been designed for more efficient operation at the architectural level.
Forecast 3: The year 2024 will underscore the importance of network visibility and reliability. The dynamic nature and changing performance of the internet will require a constant and comprehensive approach to planning and minimizing inevitable disruptions.
Today’s digital economy depends on the internet as a connecting fabric between customers, applications, and brands. As cloud solutions, SaaS, and SD-WAN become widespread in organizations, companies grapple with the complexity arising from dependence on external networks over which they have no control. Failures and other disruptions are a question of “when,” not “if.” Connectivity, security, and the assurance of delivering digital experiences regardless of the infrastructure are critically important for mitigating any disturbances in business operations.
Research conducted by the Cisco ThousandEyes Internet Intelligence team indicates two new trends, distinct from those of previous years, in terms of their size and scope. That is, with the exception of a few multi-hour outages, the condition of the internet in 2023 can be described as many more interruptions, but of a smaller scale.
Large downtime, where network or cloud disruption will have a cascading impact on dozens of other important applications or services, can be explained by a force majeure situation. On the other hand, smaller interruptions mean that the obligation to identify the source of service quality deterioration or disruptions that affect operations falls on the organizations themselves.
In a world of smaller, limited downtimes, the question, “Is the root cause on my end or someone else’s?” still requires an answer, with the addition of a practical aspect because end users want to know, and the continuity of business depends on it.
