System Capacity Calculator

Use this calculator to estimate system capacity including servers, storage and bandwidth requirements. Although the aim should be to design scalable systems regardless, back of the envelope capacity estimations help to better understand how increased demand might affect the reliability, consistency and availability of the system.

Servers

How many total users?

The entire user base (total number of possible users)

Percentage of users that use the system daily?

The % of users which are daily active users (DAU)

How many transactions does a typical user complete per session?

A transaction is a single operation but a group of HTTP requests. A single transaction might be a login operation which generates 5 HTTP requests.

How many requests does a typical transaction create?

The number of HTTP requests a typical transaction creates. A single transaction might be a login operation which generates 5 HTTP requests.

Requests per second handled by a single CPU?

The amount of requests per second a single CPU can handle. If a client waits 200ms (0.2s) for a single core CPU server to return the first byte, it can handle 5 requests per second (1s / 0.2s = 5).

How many cores in a single server?

The number of cores a single server has.

Percentage of total requests at peak hour?

The percentage of total requests that could occur within a 1 hour window. This is to consider peak times, where certain parts of the day see more traffic than usual.

Storage

Percentage of users that upload data?

This could be adding comments, liking a post or uploading photos and video.

How many uploads per day?

The average number of uploads that a typical user completes per day. This could be any data type such as adding text, images, videos or another file type. This will be used along with the average file sizes below to estimate storage requirements.

What is the most uploaded data type?

The data type users upload the most frequently. As a reference, for YouTube this would certainly be videos! This will be used with the average file sizes below to assign a weighting in estimating the storage requirements.

Text

Images

Videos

Other files like MP3, PDF, DOCX, CSV...

Percentage most uploaded type accounts for?

The % of all uploads that are of the most uploaded data type. You can adjust this weighting to amend the estimated storage.

Average text upload size in bytes?

A single character is approximately 1 byte. As a reference, a tweet on Twitter has a max length of 280 characters so requires 280 bytes. Set to 0 if no text uploaded.

Average image upload size in KB?

A low quality image (thumbnail) is ~20KB, a medium quality image (typical website image) is ~200KB, and a high quality image is ~2000KB (2MB). Set to 0 if no images uploaded.

Average video upload size in MB?

A low quality 1 min 720p video is ~256MB, a medium quality 1 min 1080p video is ~403MB and a high quality 1 min 4K video is ~2000MB (2GB). Set to 0 if no videos uploaded. Compressed video files could be lower at ~40MB.

Average file upload size in KB?

A 3 min MP3 file is ~4000KB (4MB), a Word DOCX file is ~30KB, a PDF is ~250KB and an eBook is ~3500KB (3.5MB). Set to 0 if no files uploaded.

Bandwidth

How many page views for a user per day?

The number of pages a typical user views (loads) in a day. As a reference, a social media user might view a news feed page 10 times and their profile page 10 times in a day.

Average size of a page in KB?

The transfer size of a typical web page sent to the client from the server. The average web page size is ~2MB which is ~2,000KB. You can use a tool like tools.pingdom.com to find web page size. Check as many pages as you can so the average will be more accurate.

How do I find the total size of a page in Chrome DevTools? (Opens in a new tab)

How many downloads for a user per day?

The number of downloads a typical user processes in a day. If the site has no downloadable content set as 0.

Average size of a download in MB?

The average size of downloadable content. As an example, most pages on YouTube might require a user to download a 40MB video. If the site has no downloadable content set this to 0.

Redundant factor

This will add some extra bandwidth 'padding' to the estimation to compensate for additional visits.

Formula is (Required RPS / RPS for a single server)

SERVERS REQUIRED

Servers

Formula is (Total users / 100) * Percentage of users that use the system daily

Daily Active Users

Formula is total requests per day / seconds in a day (86,400). RPS is Requests Per Second.

RPS Required

Formula is RPS for a single CPU * cores in a single server. RPS is Requests Per Second.

RPS Single Server

Formula is daily active users * transactions per session

Transactions Per Day

Formula is transactions per day * requests per transaction

Requests Per Day

At peak hour

Formula is ((total uploads per day / 100) * data type weighting) x average upload size for each data type.

DAILY STORAGE REQUIRED

Bytes

Formula is (total users / 100) * percentage of users that upload data.

Uploaders

Formula is total uploaders * uploads per day.

Daily Uploads

Formula is ((page views per day x average page size in KB * daily active users) + (total downloads per day * average download size in KB)) * redundant factor

DAILY OUTGOING DATA

Bytes

0 Bytes

Formula is same as daily storage required ((total uploads per day / 100) * data type weighting) * average upload size for data type.

Incoming Per Day

NaN

Formula is data incoming per day / seconds in a day (86,400).

Incoming Per Second

NaN

Formula is data outgoing per day / seconds in a day (86,400).

Outgoing Per Second

The server and bandwidth requirements are system aggregate calculations. You would still need to identify and test bottlenecks. These might be database connections, hard disk reading/writing, computationally intensive functions, API calls to 3rd party services. Every part of the system needs to be able to support the estimated workload.