top of page

What is Data Transfer?

Oct 23
12 min read
What is Data Transfer?

 

Data transfer is the movement of digital information between devices, systems or locations. It includes copying files to a USB drive, sending data across your home network, downloading a game and uploading a backup to cloud storage. In a usage allowance or hosting report, data transfer can also mean the quantity of data sent or received during a period. That quantity is different from the transfer rate, which describes how much data moves per second.

 

A transfer might stay entirely within your local network or cross several networks on the internet. Its completion time depends on the amount of data actually transferred, the useful rate achieved and any preparation, processing or verification the application must perform. A fast broadband package cannot remove every other constraint.

 

The Techie Broadband terminology guide connects this subject with bandwidth, storage and networking. The infographic introduces the main concepts; the worked examples below show how to use them when planning and diagnosing transfers.

 

Data transfer infographic explaining volume, rate, source and destination, limiting resources, transfer time and receipt and integrity verification

 

Data transfer, bandwidth and transfer rate

 

The first question is what the number describes. A 100 GB transfer, a 100 Mbps connection and a two-hour job measure different things. Treating them as interchangeable leads to misleading estimates and unnecessary troubleshooting.

 

Term

What it describes

Example

Transfer volume

The amount counted at a stated measurement point

100 GB of files to copy

Transfer rate

Data moved per unit of time

An application averaging 10 MB/s

Bandwidth

Available data-carrying capacity

A nominal 100 Mbps connection

Throughput

Rate actually achieved at the measured layer

A test reporting 85 Mbps

Goodput

Useful application data delivered per unit of time

Completed file bytes divided by transfer time

Latency

Delay across a path or for a response, depending on the measurement

A 20 ms round-trip time

 

Bandwidth describes capacity rather than a promise that every application will use it fully. The available capacity is shared with other traffic, and the broadband access link is only one component between source and destination.

 

When comparing results, state where the bytes were counted. A network-interface counter includes traffic that a file-copy progress bar may exclude. An application's displayed rate may represent logical file content rather than bytes currently crossing the wire, particularly when compression or reuse of existing data is involved.

 

Bits, bytes, MB and Mbps

 

Broadband rates are normally expressed in bits per second. File sizes are normally expressed in bytes. One byte contains eight bits, so the uppercase B matters: MB/s and Mbps do not describe the same rate.

 

Unit

Meaning

Exact relationship

Mbps

Megabits per second

1,000,000 bits per second

MB

Megabyte

1,000,000 bytes

GB

Gigabyte

1,000,000,000 bytes

MiB

Mebibyte

1,048,576 bytes

GiB

Gibibyte

1,073,741,824 bytes

 

NIST's explanation of decimal and binary prefixes sets out these distinctions. Software does not always label quantities clearly, so confirm whether a displayed size uses decimal or binary units before drawing conclusions from a small discrepancy.

 

To convert a decimal rate in Mbps to MB/s, divide by eight. Thus 100 Mbps corresponds to 12.5 MB/s and 1 Gbps corresponds to 125 MB/s. These are unit conversions. If the original figure is a link rate, they are not guarantees of useful file-transfer performance.

 

Similarly, 100 GiB contains about 7.4% more bytes than 100 GB. At the same useful transfer rate, it takes about 7.4% longer to transfer. That difference comes from the units, not a mysterious loss of broadband performance.

 

How does data transfer work across a network?

 

The sending application reads or generates data, represents it in a format the receiving application understands and passes it to the networking system. Network protocols carry it across links and routers. The receiving system processes the arriving data and presents it to the destination application, which may then write files, update records or perform further work.

 

A file is not necessarily sent as one network packet. Application messages, transport segments and link-layer frames have their own boundaries. The receiver's software reconstructs the content according to the protocols in use.

 

Reliable transport and application completion

 

TCP provides applications with a reliable, ordered byte stream. Sequence information, acknowledgements and retransmission help recover data lost in transit. The TCP specification in RFC 9293 describes these transport mechanisms.

 

That does not tell TCP whether the application received the intended document, wrote it durably to storage or successfully imported its contents into a database. Those are application and storage questions. The distinction explains why a network connection can finish successfully while a subsequent import or verification step fails.

 

Not every network transfer uses TCP. QUIC runs over UDP and supplies reliable streams and loss recovery itself. UDP's basic service therefore does not mean that every application built on it lacks reliable delivery. Equally, an application using raw UDP without such mechanisms cannot assume them into existence.

 

Traffic flows in both directions

 

A download predominantly carries the main payload towards you, but requests, acknowledgements and control messages also travel back. An upload reverses the main payload direction. The reverse traffic is normally a different amount, not another full copy of the file.

 

Congestion or severe delay in either direction can affect the exchange. TCP also limits outstanding data through congestion control and the receiver's advertised window. Round-trip time and the amount of data that can remain in flight help determine how effectively a flow uses available capacity. A high line rate alone does not guarantee a high rate for one connection.

 

Uploads, downloads and local transfers

 

Uploading a video or cloud backup from home depends on your upstream path. Upload speed is therefore especially relevant to creators, home-server operators and anyone sending large datasets off site. A product's headline downstream rate does not answer how quickly that upload will finish.

 

Downloading reverses the main data direction. Download speed matters for receiving software, games, media and remote files, but the sending service, route and local device still need to keep up. Full fibre describes the access technology; check the actual upstream and downstream rates rather than assuming they are identical.

 

Copying between a laptop and a NAS on the same home network can avoid the broadband connection altogether. The relevant limits may be WiFi, Ethernet, a switch, the NAS processor or storage. If an application relays the data through a cloud service, however, the apparent local task may also use the internet. Follow the actual path.

 

The same principle applies to a USB or internal-drive copy. It is still data transfer, but a broadband speed test cannot diagnose its storage-interface or disk bottleneck.

 

How to calculate data transfer time

 

Use compatible units and a rate that represents the data you are counting. For decimal gigabytes and megabits per second:

 

Transfer time in seconds = data size in GB × 8,000 ÷ rate in Mbps.

 

For decimal megabytes, use size in MB × 8 ÷ rate in Mbps. Alternatively, divide bytes by bytes per second. These are the same calculation after unit conversion.

 

The following Techie Broadband calculations assume a 100 GB file payload and a constant useful payload rate. They exclude additional preparation and finalisation time. They are theoretical examples, not measurements of a UK provider or predictions for a particular service.

 

Useful payload rate

Calculation

Time for 100 GB

20 Mbps

100 × 8,000 ÷ 20 = 40,000 seconds

11 hours 6 minutes 40 seconds

100 Mbps

100 × 8,000 ÷ 100 = 8,000 seconds

2 hours 13 minutes 20 seconds

500 Mbps

100 × 8,000 ÷ 500 = 1,600 seconds

26 minutes 40 seconds

 

If you substitute an advertised link rate for the useful payload rate, treat the result as an idealised lower bound for transmitting that uncompressed payload. Protocol overhead, sharing, retransmission and endpoint limitations can make the useful rate lower. Compression or reusing data already at the destination can reduce the amount that needs transmission instead.

 

Do not subtract an arbitrary overhead percentage from a rate already measured using useful file bytes. The effects that reduced that measured rate are already reflected in it. Match the numerator and denominator before making an adjustment.

 

Will a 100 GB backup fit into four hours?

 

Working backwards from a deadline is often more useful than asking how fast a connection is. To send 100 GB in four hours requires an average useful payload rate of 100 × 8,000 ÷ 14,400 = about 55.6 Mbps. This leaves no time allowance for scanning, outages or final verification.

 

At a constant 20 Mbps of useful payload throughput, four hours carries 36 GB. The full 100 GB cannot fit into that window under those assumptions. The options include a longer window, a higher achieved upstream rate or reducing the required transfer through a suitable incremental method.

 

If only 4 GB genuinely needs sending on a later run, the same 20 Mbps rate gives 1,600 seconds, or 26 minutes 40 seconds, for that payload. This does not mean a backup tool necessarily sends only the number of bytes you edited. Its change detection, chunking, compression and file formats determine the actual transfer.

 

Allow margin based on the workload you observe. A plan that just fits mathematically can still miss the deadline when another household upload starts or the destination spends time processing the result.

 

Why thousands of small files can take longer

 

Data volume is only part of a file-copy workload. Opening files, reading metadata, checking permissions, creating destination entries and receiving application responses can add repeated work. This is why one large archive and a directory of many small files need not finish at the same time even when their total file sizes match.

 

Consider a deliberately simplified model: 20,000 files contain 1 GB in total, and each file adds 20 ms of serial work that cannot overlap with payload transfer. The payload takes 80 seconds at 100 Mbps useful throughput. The serial work adds 20,000 × 0.02 = 400 seconds, giving 480 seconds, or eight minutes, in total.

 

The 20 ms is an assumption, not a measured disk delay or an inevitable network round trip per file. Real software can batch, pipeline or parallelise work. The example shows why improving the link rate might produce a disappointing improvement when much of the job is spent doing something else.

 

Why transferred bytes can differ from file size

 

There are several quantities worth keeping separate: the source's logical file size, bytes the application actually sends, traffic counted on a network interface and any volume counted by a service for usage or billing. They need not match.

 

  • Compression can reduce transmitted content when the data is compressible; already compressed or encrypted content may offer little further reduction.

  • Incremental or delta methods can reuse suitable content already at the destination, depending on the tool and configuration.

  • Protocol headers, control traffic and retransmissions can add network bytes beyond the useful payload.

  • Retries or restarting a failed upload can repeat part or all of a transfer.

  • Metadata and application processing can add work without adding a proportionate amount of file content.

 

For example, rsync can use a delta-transfer algorithm to send differences relative to a destination copy. That is a tool-specific mechanism, not a guarantee that every synchronisation service transfers only changed bytes. Compression and delta transfer can also consume CPU or storage I/O, shifting the constraint elsewhere.

 

Data transfer allowances and cloud accounting

 

A monthly quantity such as GB or TB is an amount over a period. It is not a connection rate. A large allowance cannot make one transfer faster, and a high transfer rate says nothing by itself about the volume permitted or charged.

 

Check which traffic a service counts, in which direction, over which boundary and for which reporting period. A cloud platform may distinguish incoming traffic, outgoing traffic and transfers between locations. Check that service's current terms before calculating costs. This article does not assume a particular provider's charging rules or quote a price.

 

An unlimited volume allowance is also not a promise of unlimited instantaneous capacity. Capacity, acceptable-use conditions, service limits and storage allocation are separate questions.

 

Security, interruption and verification

 

For internet transfers, use an appropriate authenticated, encrypted protocol and verify that you are connecting to the intended service. TLS protects data in transit for applications such as HTTPS. It does not automatically secure the destination account, encrypt every stored copy or make an unsuitable recipient safe.

 

Permissions matter separately from transport security. Being able to connect to a server does not imply permission to read a particular file or write into a particular directory. A denied transfer may be an access-policy problem rather than a broadband fault.

 

Can an interrupted transfer resume?

 

Sometimes. The client, server and operation must support it, and the partial content must still belong to the intended version. HTTP provides optional range requests for retrieving portions of a representation. Its conditional mechanisms can help avoid combining an old partial copy with changed content.

 

The HTTP semantics standard specifies range and conditional request behaviour. It does not guarantee that every download endpoint supports resuming, or that an upload operation supports it in the same way. A cloud service may use a separate multipart or resumable-upload mechanism.

 

A progress bar returning to zero does not prove that all network data was lost. The application may be repeating preparation or processing, or may simply not support restarting from its earlier position. Read its log or status before estimating how many bytes need sending again.

 

What proves that a transfer is complete?

 

Start with the application's success result, then use verification appropriate to the task. For important file copies, compare the expected file set and sizes and, where appropriate, cryptographic hashes calculated from the intended source and destination contents. A matching name or size alone cannot prove that every byte matches.

 

Obtain any published expected hash through a trusted route. A hash supplied beside an untrusted download is not independent evidence that its publisher is trustworthy. Hashes compare content; they do not grant permission or establish that software is safe to run.

 

For databases or changing datasets, byte comparison alone may not establish a consistent migration. Use the application's supported snapshot, export or migration process and check the imported result. Where storage durability matters, understand what the destination's success acknowledgement actually promises.

 

Finally, a copied file is not automatically a useful backup. Retention, independent recovery points and an actual restore test determine whether you can recover the required state. Synchronisation can propagate unwanted changes as efficiently as wanted ones.

 

Diagnosing slow or failed data transfers

 

Use measured throughput from the actual workload alongside other observations. A single speed test can help characterise a path to its test server, but it cannot prove that a different destination, application or storage system can sustain the same rate.

 

Observation

Possible explanation

Useful next check

Local NAS copy is slow

WiFi, LAN link, NAS CPU or storage limit

Compare a wired copy and endpoint resource usage

Downloads are fast but uploads are slow

Lower upstream capacity, competing uploads or receiving-service limit

Measure the upload direction and inspect other outgoing traffic

One remote service is slow

Destination load, account limit or route-specific problem

Compare another suitable destination and the service's status

Many small files are slow

Metadata, serial application work or storage operations

Compare with one large non-sensitive test file

Network traffic finishes but the job continues

Hashing, writing, indexing, importing or other finalisation

Inspect the application's current stage and logs

Transfer fails immediately

Authentication, permissions, path or quota problem

Read the exact error before changing network settings

 

An effective diagnostic sequence changes one variable at a time:

 

  1. Record the source, destination, direction, file set, units and application stage.

  2. Measure elapsed time and useful completed bytes over a representative interval.

  3. Establish whether the path is local, internet-facing or relayed through a service.

  4. Compare wired connectivity where practical and account for competing traffic.

  5. Inspect sender and receiver CPU, storage activity, permissions, space and application limits.

  6. Retest a comparable workload after changing the suspected constraint, then verify the destination result.

 

Avoid treating every low rate as ISP throttling or every failure as packet loss. A fixed application rate limit, full destination disk and overloaded WiFi connection require different solutions. Repeating an unchanged test gives more data only if you are measuring a useful difference, such as time of day or competing load.

 

Improving transfer performance sensibly

 

First identify which part of the job consumes time. If payload transfer is the constraint, improve that path or reduce the bytes required. If scanning or finalisation dominates, focus on the application and endpoints.

 

Useful changes can include wired networking, scheduling bulk uploads away from busy periods, using a supported incremental method and choosing an appropriate amount of concurrency. More simultaneous transfers share finite resources and can overload storage, CPU or queues. They do not multiply the connection's available capacity.

 

Archiving many small files can reduce per-file operations, but it adds archive creation and extraction and may change incremental-update behaviour. Compression is worth testing with the actual data rather than assuming a fixed saving. Preserve required permissions, timestamps and other metadata when selecting a copy method.

 

For a one-off migration, include preparation, validation and cutover time in the plan. Data transfer is one stage of the outcome. The real test is whether the intended system can use the correct data afterwards.

 

Key takeaway

 

Ask three questions: how much data needs to move, what useful rate can this path sustain, and what must happen before the result is usable? Those answers give a stronger basis for planning than a headline broadband speed alone.

 

  • Volume, rate and elapsed time describe different properties.

  • File bytes, network traffic and service-accounting totals can differ.

  • Local hardware, application behaviour and destination processing can limit the job.

  • Successful delivery should be followed by verification suited to the data and its purpose.

 

Frequently asked questions

 

Is data transfer the same as download speed?

 

No. Data transfer describes moving information or the quantity moved. Download speed describes the rate towards your device. Uploads and local copies are also data transfers.

 

How long does it take to transfer 100 GB?

 

At a constant useful payload rate of 100 Mbps, 100 decimal GB takes 8,000 seconds, or 2 hours 13 minutes 20 seconds. Preparation, finalisation and interruptions add time, while compression or reusing existing content can change how many bytes need transmission.

 

Why is my transfer speed shown in MB/s rather than Mbps?

 

Applications often report bytes per second while broadband rates use bits per second. One byte contains eight bits, so 12.5 decimal MB/s equals 100 Mbps. Check whether the application instead uses binary MiB/s before comparing figures.

 

Does transferring files at home use broadband data?

 

A direct transfer between devices on the same local network can stay off the broadband connection. A transfer to remote storage or through a cloud relay uses an internet path. The application's actual route determines which links carry the data.

 

Does an encrypted transfer mean the destination file is safe?

 

Encryption protects data over the connection under the protocol's security model. It does not prove that the recipient is appropriate, the destination account is secure or the stored result is a usable backup. Verify the destination and content separately.

 

Does data transfer mean deleting the original?

 

Not necessarily. A copy usually leaves the source intact. A move operation may remove the original after copying, while synchronisation or migration can follow other rules. Confirm the tool's behaviour before transferring important data.

 
 
UK Techie Broadband Providers
bottom of page