[1/3] Google's AI-powered next-generation global network: Built for the Gemini era (Category: Infrastructure)
[1/3] Google's AI-powered next-generation global network: Built for the Gemini era (Rubric #Infrastructure)
Read it. story From Google on the evolution of their network by Bikash Koley, VP on Wide Web and Infrastructure. The main idea of the article is to show the evolution of Google’s private global network and new design principles designed to meet the rapidly growing needs of the AI era. (Advertise the availability of this Google network to GCP customers as a product CloudWAN).
These are the eras of Google’s network architecture. 🌐 Internet era (2000-e) The focus was on fast and reliable search results, Gmail, maps, etc. To do this, Google built its own data centers and communication channels, and also invented technologies that allowed to scale the network: a private highway, the first software-defined WAN B4traffic controller Orion and datacenter switch Jupiter 📱 Streaming era (ending 2000- x) With the growth of YouTube and streaming video, Google has adapted the network for video streaming – reduced delays and jitters thanks to the development of its CDN. (Google Global Cache - caching nodes for telecom operators) and new data transmission protocols (Espresso, QUIC, TCP BBR etc.) 💭 Cloud era (2010-e) Then came the rapid growth of cloud services, which required increased reliability, customer isolation and network security. Google responded by implementing SDN (software-defined networks) everywhere: from a virtual data center network Andromeda Up to the new GRPC protocol and traffic protection systems (PSP, Swift etc.).
Now the Google network is very large.
- 2 Millions of miles of fiber optics, investments in 33 Submarine cables across oceans that connect compute infrastructure
- 200+ Point of Presence nodes, 3000+ CDN locations around the world, 42 cloud regions, 127 zone
In continuation I will tell you how and where Google plans to develop its networks and compare it with the approach from the banned in Russia Meta.
#Software #DevOps #Architecture #Economics #DistributedSystem #SystemDesign