Requirements

Let’s sum up what we want our monitoring system to do for us:

• Monitor critical local processes on a server for crashes.
• Monitor any anomalies in the use of CPU, memory, disk, or network bandwidth by a process on a server.
• Monitor overall server health, such as CPU …

We’ll discuss the core components of our monitoring system, identify the shortcomings of our design, and improve the design to fulfill our requirements.

Storage

We’ll use time-series databases to save the data locally on the server where our monitoring service is running. Then, we’ll integrate it with …

Large data centers have millions of servers, and visualizing the health data for all of them is challenging. An important aspect of monitoring a fleet of servers is to know which ones are alive and which ones are offline. A modern data center can house many thousands of servers in …

Motivation

There can be millions of events happening per second in a large distributed system. Commenting on a post on Facebook, sharing a Tweet, and posting a picture on Instagram are just a few examples of such events. We need a mechanism to distinguish these events from each other. One …

In the previous lesson, we saw how unique identifiers are used in many systems, for example, to label tweets, uploaded videos, or track execution flow across services. In this lesson, we will build on that by exploring different ways to design a system that generates such unique IDs. We will …

Causality

In the previous lesson, we generated unique IDs to differentiate between various events. Apart from having unique identifiers for events, we’re also interested in finding the sequence of these events. Let’s consider an example where Peter and John are two Twitter users. John posts a comment (event …

Problem statement

Let’s start with a question: If millions of users worldwide use our data-intensive applications, and our service is deployed in a single data center to serve the users’ requests, what possible problems can arise?

The following problems can arise:

• High latency: The user-perceived latency will be high …

A content delivery network (CDN) is a distributed system of servers placed in many geographic locations to improve the speed and reliability of delivering online content. Rather than relying on a single origin server that holds the original version of the data, a CDN uses a network of proxy servers …

CDN design

We’ll explain our CDN design in two phases. In the first phase, we’ll cover the components that comprise a CDN. By the end of this phase, we’ll understand why we need a specific component. In the second phase, we’ll explore the workflow by explaining …

Introduction to key-value stores

Key-value stores are distributed hash tables (DHTs). A key is generated by the hash function and should be unique. In a key-value store, a key binds to a specific value and doesn’t assume anything about the structure of the value. A value can be a …