Object storage vs block storage vs file storage.

Cloud providers sell storage in three kinds: file, block and object. They differ in how the data is organised and how software reaches it. Each suits different work, and they are priced differently, so it helps to know which one a quote is talking about.

File storage

File storage keeps data as files inside folders, and folders inside other folders. It is the arrangement everyone knows from their own computer. In the cloud it is offered as a shared drive that many computers can open at once, using the same network protocols as an office file server.

Microsoft suggests its file service for replacing or adding to an office file server or NAS, and for moving an application that already expects a shared drive. If staff need to open, edit and save the same documents, this is the kind you want.

Block storage

Block storage cuts data into fixed-size blocks, each with its own identifier, and reads or writes them one at a time. To the server it is attached to, it looks like an ordinary hard disk. Microsoft describes its version as a physical disk in a server, but virtual.

Because a change touches only the blocks affected, block storage is quick at many small reads and writes. That is why it sits under databases and under the system disk of every virtual server. It has two limits. A disk is normally used by the server it is attached to and not reached from outside it. And you pay for the size you set aside, whether or not it is full.

Object storage

Object storage keeps each item whole, as an object. An object carries a unique identifier and a set of labels, called metadata, that describe it. There are no real folders. Objects sit in one flat pool and are fetched by identifier, usually by a program working over the web.

This arrangement grows almost without limit and is charged by what is stored. Amazon S3, Azure Blob Storage and Google Cloud Storage are all object storage. Typical uses are backups, archives, and the images, video and documents behind a website or an app.

The trade-off is editing. Google notes that an object cannot be modified once it is created: to change it, the whole object is uploaded again. That makes object storage a poor home for a database or for a spreadsheet five people have open.

The real difference

• File is organised by folder path and shared across a network. People work in it directly.
• Block is organised by block number and attached to one server. Software that needs speed works in it.
• Object is organised by identifier and label and reached over the web. Things are put there to be kept.

How to choose

In most cases the software chooses. A virtual server needs block storage for its disk. An accounting package that expects a shared drive needs file storage. A backup service will usually ask for object storage.

A small business often ends up with all three without having picked any of them. What is worth doing is asking, for each line on a cloud quote, which kind it is and what it holds. Then check two things: that nothing people need to edit together has been put in object storage, and that the block storage set aside for each server is close to what it uses.