Koley says the whole idea of Andromeda is to take network functions and virtualize them -- this includes functions like routing packets, load balancing, filters, access control lists, DDoS mitigation and potentially deep packet inspection. In a traditional network, you would have specialized boxes for these functions. Google's Andromeda virutalizes all of that and makes it available as simple APIs, providing progammable access to the same infrastructure that runs all of Google's services. Koley says this essentially delivers the "somewhat magical scaling" that would be difficult to achieve any other way.
00:03 -What is Andromeda?
04;56 - Since Andromeda supports virtualized network functions, will Google be offering hosted network services like hosted firewalls?
05:25 - To orchestrate the Google infrastructure, Andromeda must deliver performance and scalability. How are these attained?
06:52 - Is Andromeda an open platform for developers?
Watch 8-minute video: http://youtu.be/wpin6GKpDm8
In April 2014, Google announced that its "Andromeda" network virtualization stack is now powering two of its Google Compute Engine zones: us-central1-b and europe-west1-a. The company will be migrating its other data centers to Andromeda in the coming months.
Google describes Andromeda as a Software Defined Networking (SDN)-based substrate that serves as the orchestration point for provisioning, configuring, and managing virtual networks and in-network packet processing. The goal is to expose the raw performance of the underlying network while simultaneously exposing network function virtualization (NFV), including distributed denial of service (DDoS) protection, transparent service load balancing, access control lists, and firewalls.
Google reports a significant performance gain in throughput for customers in the data centers running Andromeda, with TCP throughput for 200 streams soaring from under 2 Gbps (baseline) to 5 Gbps (powered by Andromeda).