Cloud storage explained simply: Understanding the basics

Applications generate a significant amount of data: user files, images, videos, backups, activity logs, and analytics data. A few years ago, a single physical server was enough for many projects. But as requirements have grown, managing dozens of terabytes of data yourself has quickly become complex and expensive.

Cloud storage addresses this challenge by allowing data to be stored on remote infrastructure managed by a cloud provider while offering excellent scalability.

In this article, we'll look at how cloud storage works, the different types of storage available, and their main characteristics.

What is cloud storage?

Cloud storage consists of storing data on remote servers accessible over the Internet and hosted in a cloud provider's data centers.

The main providers offer their own storage solutions:

• Amazon S3 for Amazon Web Services (it has become such a standard that it hardly needs an introduction).
• Google Cloud Storage for Google Cloud.
• Azure Blob Storage for Microsoft Azure.

Users do not need to manage disks, servers, or maintenance themselves. The cloud provider is responsible for the physical infrastructure, data replication, and part of the service availability.

Cloud storage can be used for many purposes, such as storing files uploaded by users, hosting images or videos, keeping backups, storing logs, and more.

The three main types of cloud storage

Not all cloud storage works the same way. The choice mainly depends on the type of data and how it will be used.

Object Storage

Object storage is one of the most widely used storage models. It is designed to store large amounts of unstructured data such as images, videos, files, backups, logs, and analytics data.

Unlike a traditional file system, object storage does not organize data into a hierarchy of directories. Data is stored as independent objects inside containers generally called buckets.

An object is made up of three main components:

• The data: the binary content being stored (image, archive, video, document, etc.).
• Metadata: information associated with the object, such as its MIME type, creation date, size, or custom attributes.
• A unique key: an identifier used to locate the object inside the bucket.

For example, an object stored in Amazon S3 may have a key such as:

users/123/avatar.jpg

Although this looks like a traditional file path, users/123/ is not actually a directory. It is simply a prefix used to organize and filter objects.

Data is generally accessed through HTTP REST APIs. For example, an application may send the following request to retrieve a specific object:

GET /users/123/avatar.jpg

One of the biggest advantages of object storage is its ability to handle multiple petabytes of data using an architecture designed for high durability. On the other hand, it is generally not suitable for applications that require frequent random modifications to parts of a file, such as certain databases or systems requiring traditional disk access.

File Storage

File storage is the storage model that most closely resembles the traditional file system used on a server. Data is organized into a hierarchy of directories and files. Unlike object storage, applications can browse this structure directly and perform standard file operations:

• Create files.
• Read and write data.
• Modify files.
• Delete files.
• Manage permissions.

In a cloud environment, file storage is generally accessible over the network. Multiple servers or virtual machines can mount the same storage space simultaneously and work with the same files.

The most common protocols are:

• NFS (Network File System): widely used in Linux and Unix environments.
• SMB (Server Message Block): mainly used in Windows environments, especially for file sharing within organizations.

Cloud providers offer managed file storage services such as Amazon EFS, Azure Files, and Google Filestore. These services provide a shared file system without requiring users to manage storage servers themselves.

File storage is not always the best choice. For very large volumes of unstructured data or files mainly accessed through an API, object storage is often more appropriate. On the other hand, applications requiring the performance of a local disk are generally better suited to block storage.

Block Storage

Block storage provides a raw storage volume presented to a machine as if it were a physical disk. The operating system can then format it with a file system (ext4, NTFS, etc.) or allow an application to access it directly.

Unlike object or file storage, block storage does not manage file organization. It is mainly optimized for applications requiring high performance and low latency.

Typical use cases include:

• A virtual machine's system disk.
• Databases requiring fast data access.
• Applications with high disk performance requirements.

Cloud services such as Amazon EBS, Azure Managed Disks, and Google Persistent Disk are based on this model.

The different storage classes

Cloud providers offer several storage classes to adapt pricing according to how frequently data is accessed.

Standard Storage

This class is intended for data accessed frequently. Examples include:

• Files used by a web application.
• Frequently displayed images.
• Data used every day.

It generally provides fast access to data but has a higher storage cost than archive-oriented classes.

Nearline Storage

This class is designed for data accessed occasionally. Examples include:

• Recent archives.
• Rarely used backups.
• Historical data.

Storage costs are generally lower, although some operations may be more expensive.

Cold Storage

Cold storage is intended for data that is rarely accessed. Examples include:

• Older backups.
• Regulatory data.
• Internal archives.

Its main goal is to reduce storage costs.

Archive Storage

This class is intended for data kept for several years. Examples include:

• Legal documents.
• Long-term backups.
• Historical data.

Retrieving data may take longer and incur additional charges.

Costs to consider

The cost of cloud storage is not limited to the amount of data stored. Several factors can affect the final bill:

• The number of operations performed on the data (read, write, delete).
• The storage class being used.
• Retrieval costs for archived data.
• Data transfer outside the cloud provider (egress).

For example, an application storing a large number of videos may generate higher transfer costs than storage costs.

Cloud can seem almost magical at first, with resources available in seconds without having to manage any physical infrastructure. However, this simplicity should not hide the reality of cloud pricing: poor configuration or an uncontrolled deployment can quickly make costs soar.

Conclusion

Cloud storage is much more than simply replacing a hard drive with a remote server. It offers different storage models designed for different needs. Thanks to its scalability, resilience, and numerous security mechanisms, it has become an essential part of today's IT infrastructures.

However, using cloud storage efficiently requires choosing the right storage model, understanding the associated costs, and configuring access properly to benefit from its advantages without creating new problems.