The benefits of bare-metal commodity hardware in the CDN infrastructure are numerous, and will be even more in the coming years, but at the same time so will the advancements in artificial intelligence and deep learning. Today’s dumb CDN PoPs will become highly intelligent PoPs in the next few years, where the software stack at each PoP will be able to analyze massive volumes of web traffic, click patterns and customer behavior, and make decisions on the fly about how to process, store and serve each piece content to whatever device is destined for. Deep learning, which is another way of saying machine learning is already highly visible in the cyber security world, where software platforms like FireEye and Bromium deploy machine learning, correlation, big data and advanced algorithms to make decisions on the fly. In addition, this technology is slowly making its way into the world of caching, streaming and delivery.
Recently, Facebook Artificial Intelligence Research (FAIR) open sourced it software libraries to the public, and they have gone on record stating their software is ahead of the competing products. Google acquired DeepMind for $400M to $500M, where its machine learning technology is being used in eCommerce simulations. Although a lot of AI research is being done in academia, for the purpose of furthering scientific research, the technology is showing in commercial products. AI research touches upon the following areas: machine learning, neural networks, image processing, parallel processing, networking, and so on. Takeaway: Its only a matter of time before it goes from pie-in-the-sky to reality, where CDNs create a product and business model from it.
Artificial Intelligent CDN Tidbits
- Intelligent CDN PoPs acts like self running CDN units
- Dynamic request are served from their “spheres” eliminating global round trips for dynamic content
- Diminished use of middle mile accelerations technologies since global round trips are minimized
- Deep Learning Cyber Security:Â PoPs determine if request is malware or bot, and then takes immediate action through machine learning algorithms to detonate
- Performance on Demand: PoPs will monitor the compute capacity of all the servers in the racks, and assign under-utilized capacity to over-utilized processes