AI, 5G and 400GE – The Perfect Storm
2018-09-25 | 6 min read
Artificial intelligence (AI) is a hot topic these days, but it is by no means a new topic. The concept of AI has been around since the 1950s. In 1955, the term ‘artificial intelligence' appeared in A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. AI has been in and out of fashion so to speak ever since. Until now, when it seems that we finally have the technology and computing power in place for AI really to flourish.
One of the main reasons that AI did not take off decades ago is that the sheer amount of computing power and network speed it required simply did not exist back then. Today, 5G and 400GE are two game-changing technologies that will simultaneously both enable AI, as well as depend upon it. While the exact speed of 5G is yet to be determined, early trials have been promising, with some estimates that 5G download speeds will be 1,000 times faster than 4G. Latency speed, the time it takes for the network to respond to a request, is another critical factor to consider. We expect 5G latency speeds of less than ten milliseconds.
100GE is widely deployed in data centers around the world today, but 400GE links will be the next step to increase network bandwidth for 5G-capable data centers. Faster networking speeds require faster memory and faster serial bus communications. Peripheral Component Interconnect Express (PCIe®) expansion bus speeds are evolving from PCIe 4.0 to PCIe 5.0 to support these increased speeds. The same is true for memory, as double date rate (DDR) memory evolves from DDR 4.0 to DDR 5.0. All these technologies, 5G, 400GE, PCIe 5.0, and DDR 5.0 are creating the perfect storm to enable AI to take off finally.
The relationship between AI and the data center is symbiotic. Many believe that the success of next-generation data centers is dependent upon AI. Gartner predicts that "By 2020, 30 percent of data centers that fail to apply AI and machine learning effectively in support of enterprise business will cease to be operationally and economically viable." Intelligent data centers will be possible when AI is paired with software-defined networking (SDN) and network functions virtualization (NFV). NFV is a network architecture concept that automates entire classes of network node functions into building blocks that may connect, or chain together, to create communication services. SDN enables network management and control through programmable, centrally controlled applications. AI can optimize virtual network configurations and proactively distribute workloads to ensure data center efficiency. The success of 5G depends upon an AI-based network infrastructure for 5G network capacity planning and operation.
Data center infrastructure management (DCIM) will also use AI. DCIM solutions provide a holistic view of the modern data center by integrating information technology (IT) and building facility management. Next-generation data centers will use AI for DCIM with a highly automated infrastructure in which applications and data will be intelligently deployed and delivered to optimize data center operators’ top 3 concerns: cost, footprint, and power. An example is Google’s DeepMind, which applied machine learning to reduce the amount of energy used for cooling in Google data centers by up to 40 percent. Google has used AI to enhance many of its applications including the Google search engine, Gmail, and YouTube.
Top technology companies will continue to invest in AI research. Several other high-profile AI projects are underway including Facebook’s new AI lab that just celebrated its 1-year anniversary at my alma mater, McGill University, and $2 billion AI research funding recently announced by Defense Advanced Research Projects Agency (DARPA). While AI applications can undoubtedly improve our lives for the greater good, its power can also be used to the detriment of humanity. It is interesting to note that several of the top AI researchers, including Elon Musk, and DeepMind Founders have pledged not to use AI to develop lethal autonomous weapons.
With 5G and 400GE on the horizon, the possibilities of future AI-enabled applications are endless. Keysight’s data center infrastructure solutions can accelerate your 5G and 400GE innovations, paving the road for new AI applications in the future.