RAS: Continuously Optimized Region-Wide Datacenter Allocation

Facebook provides some of the popular social applications. That provides various functionalities and is used by billions of users every day. Supporting these applications require highly scalable and sophisticated backend services running on many data centers, e.g., Key-Value Stores, ML Inference, Stream Processing, Databases, etc.

RAS in the software stack

Let’s have a look at how RAS fits into the software stack.

1. First, we have our user-facing products

2. The second layer is of backend applications (Databases, ML Inference, Stream processing, Other Applications)

3. The infrastructure is divided into three main layers

   • On the top, we have Shard Management. It helps scale and operate backend services

   • The second layer is container management which manages containers on top of servers.

   • The third layer is hardware management. How does RAS manage hardware resources within a region?

Today, Facebook operates in many regions around the globe. Let’s take a closer look at the organization of a region.

Each data center region is composed of several data center buildings. Each of them is connected via high bandwidth and low latency network. Each data center building is composed of Failure Domains called the Main Switch Board (MSB). Each MSB is designed to fail independently. Each MSB is composed of tens of thousands of servers.

The first ...