What is Cloud Sync?

Cloud sync, short for cloud synchronisation, keeps selected data aligned between devices and an online service. A sync client detects changes, transfers the required data and updates other connected devices according to the service's rules. Those changes can include new files, edits, renames and deletions. Sync makes it easier to work across computers and phones, but synchronisation alone does not guarantee that an older, recoverable copy remains available.
A file appearing in a cloud folder is not proof that its contents are stored on your device, fully uploaded or available everywhere else. Understanding which of those states applies is more useful than simply knowing that sync is switched on.
The Techie Broadband terminology guide connects cloud sync with storage, networking and broadband performance. The infographic shows its main responsibilities and the distinction between keeping working files aligned and protecting a recoverable copy.

How does cloud sync work?
Imagine editing a document on a laptop, then opening it on a second computer. In a typical cloud-backed file service, the first client discovers the edit, identifies the affected file and sends the required content and metadata to the service. The second client learns about the change and retrieves the updated content when its settings require it.
The cloud service is the meeting point, so the two computers do not necessarily need to be online together. The source must finish uploading before the other device can retrieve that version. A client that is offline, asleep, paused or signed into a different account will not necessarily show the latest state immediately.
The main stages are:
Define the scope: which account, folders, shared spaces and file types participate.
Detect changes using operating-system notifications, scanning, service notifications or polling, depending on the implementation.
Compare local and remote state using metadata, version information or content fingerprints.
Transfer the content needed to create or update the destination copy.
Record the new state and apply the relevant changes to other participating devices.
Report pending work, errors, conflicts and local availability so the user can verify the result.
That is a conceptual model, not a single protocol followed by every supplier. Some systems transfer complete files, some reuse unchanged blocks and some synchronise application records rather than files. Notifications and metadata can arrive before the full payload is downloaded.
Syncthing's technical documentation provides a useful implementation example: it describes filesystem watching, scans, content hashes, blocks and a local index of known versions. Syncthing itself synchronises between devices rather than requiring a central cloud-storage account. The mechanisms help explain the problem without making every cloud client identical to Syncthing.
Generic cloud sync and the Synology product
“Cloud sync” is a general term. Synology also has a NAS application called Cloud Sync, which connects supported cloud services with selected NAS folders. Its capabilities depend on the application, cloud destination and configuration. It is one implementation of the concept, not the definition of cloud synchronisation.
Contacts, calendars, browser settings and application data can also synchronise through online services. Their conflict and recovery behaviour may differ from ordinary folder sync. A photo library or collaborative document is not necessarily represented by a straightforward downloadable file with all its metadata intact.
Two-way sync, one-way sync and local copies
There are two separate choices: which changes should travel between locations, and how much file content should occupy local storage. These choices are often confused.
Setting | What it controls | What to check |
|---|---|---|
Two-way sync | Changes from either side can affect the other | Conflict, rename and deletion rules |
One-way sync | Changes flow in a configured direction | Whether destination-only files or source deletions are removed |
Selective sync | Which folders or items participate on a device | Excluded items will not necessarily be present locally |
Full local copy or mirroring | Selected file contents are retained locally as well as remotely | Disk space, successful download and application compatibility |
On-demand or streaming access | Some visible files are downloaded when needed | Offline availability, cache behaviour and client requirements |
A one-way task is not automatically a safe backup. A mirror can remove destination files when the source removes them, while another copy task might retain those destination files. Read the actual task policy rather than infer deletion behaviour from an arrow or the word “sync”.
Google Drive for desktop documents a choice between streaming and mirroring for My Drive. Mirroring keeps a local copy; streaming primarily retains data in the cloud and provides local content as needed, with an option to make files available offline. Other folders, shared drives and operating systems have their own rules. These are storage and access choices, not promises that every connected device will hold every file.
Online-only does not mean downloaded
An online-only file can have a visible name, size and thumbnail while its full contents remain remote. Opening it may trigger a download. Before travelling or working through an outage, explicitly retain the required files locally and wait for completion. Then disconnect and test opening them in the application you intend to use.
Microsoft's OneDrive Files On-Demand documentation distinguishes online-only files, locally available files and files marked to stay on the device. Its “Free up space” action makes content online-only without deleting the cloud file. Ordinary deletion is a different operation and can remove the file from OneDrive and other devices.
Do not use the Delete key as a substitute for freeing local space. Equally, do not assume a locally cached file is permanently pinned for offline use. Check the client's status and retention setting for that item.
Is cloud sync a backup?
Cloud sync can protect access after a device fails: if the upload completed and the account remains accessible, another device may retrieve the cloud copy. That is useful redundancy. It is weaker protection against an unwanted change that the system correctly synchronises everywhere.
Capability | Main purpose | Recovery question |
|---|---|---|
Synchronisation | Keep participating working data aligned | Will an unwanted edit or deletion propagate? |
Version history or recycle bin | Retain some previous or deleted states | Which items, retention period and account conditions apply? |
Separate backup | Preserve recoverable data under a defined policy | Can the required earlier state survive loss or compromise of the working copy? |
Archive | Retain information for a specified long-term purpose | Are the contents complete, readable and protected for that period? |
Some cloud products combine these capabilities. The distinction is the recovery behaviour, not whether the supplier uses the word “backup”. Check which data is included, how old a version can be restored, whether deletions can be permanent and who can erase the recovery copies.
Apple's iCloud Drive guidance, for example, says deleting a file in synchronised Desktop and Documents folders removes it across devices using the same Apple Account. A separate “Remove Download” action frees local storage without deleting the cloud file. The same visible folder can therefore support two very different user intentions.
Ransomware can create another unwanted change: encrypted or corrupted working files may be synchronised like legitimate edits. Recovery then depends on unaffected devices, suitable retained versions or separately protected backups. A second folder controlled by the same compromised account is not automatically independent protection.
The NCSC's guidance on backing up data, reviewed in July 2026, covers both online storage and external devices. It recommends keeping external backup devices disconnected when not in use, protecting online backups with two-step verification and checking that you can restore the important data. The practical test is recoverability, rather than the number of cloud icons on your screen.
What happens when you edit files offline?
If a file is available locally and its application supports offline work, you can edit it without an active internet connection. The client must later reconcile those changes with the remote state. A file that only exists as an online-only placeholder cannot be opened offline merely because its name is visible.
If another device edits the same file before yours reconnects, neither version necessarily contains the other's work. A service might keep a conflict copy, prefer a version or use application-aware merging. Real-time coauthoring in a supported document format can handle edits differently from synchronising an arbitrary binary file.
Techie Broadband analysis: why the previous state matters
We checked an original six-case state model using Python 3.12.14. It is an explanatory model, not a test of OneDrive, iCloud, Google Drive or any production sync engine. A, B and C represent different file contents; “absent” means no file exists at that path. The baseline is the last state both sides had agreed.
Example | Baseline / local / cloud | Illustrative decision |
|---|---|---|
budget.csv | A / B / C | Both changed differently: preserve both for review |
notes.txt | A / B / A | Only local changed: upload local content |
photo.jpg | A / A / B | Only cloud changed: download cloud content |
old.txt | A / absent / A | Local deletion: propagate only under the intended deletion policy |
offline.txt | absent / A / absent | New local file: upload it |
new-name.txt | absent / A / B | Different new files share a name: preserve both for review |
The model produced two uploads, one download, one conditional deletion and two conflicts. The important insight is that absence alone does not tell you whether a file is new somewhere else or was deliberately deleted here. The agreed baseline supplies that context.
A delete on one side and an edit on the other also needs conflict handling. The model preserves that situation for review rather than silently treating one intention as correct. It does not resolve renames, permissions, simultaneous writers or every application-specific change.
Modification time is not the same thing as knowing which contents the user wants. Even accurate clocks cannot decide whether one person's later edit should discard another person's independent work. When a conflict appears, preserve the alternatives outside the active sync tree, compare them and resolve deliberately. Avoid repeatedly renaming and deleting copies while several clients are still applying changes.
Why the first sync takes longer
An initial sync may need to enumerate folders, read files, calculate fingerprints and transfer a large starting dataset. Later runs can often work from a smaller set of changes. That does not mean every edited file requires only a few changed bytes to cross the network.
Delta or block-based transfer reuses unchanged content where the implementation supports it. File formats, compression, encryption and how an application rewrites data can all affect how much content is actually reusable. Reading or hashing a large file may still consume local CPU and disk time even when the network payload is small.
There is a concrete product difference here. Synology's Cloud Sync specification for DSM 7.4 says that Cloud Sync does not perform incremental updates and re-downloads a modified file even if only part changed. That does not describe every Synology application or every cloud service. Syncthing, by contrast, documents comparing blocks and obtaining the blocks required for the desired version.
Techie Broadband calculation: a small edit can still mean a large transfer
Assume a 10 GB decimal file changes and an idealised delta transfer would need 10 MB of payload. Sending the full file moves 10,000 MB rather than 10 MB, a factor of 1,000. At a constant 20 Mbps payload rate, those amounts take 4,000 seconds, or 1 hour 6 minutes 40 seconds, versus four seconds.
This is a theoretical comparison of assumed transferred amounts, not a measured product benchmark or a promise of a four-second sync. Real block boundaries, metadata, scanning, encryption and publication add work. The useful question is how much data the actual client transfers for your files, not how small the visible edit looks.
How broadband affects cloud sync
When files leave your home or office, upload speed matters. When another device fetches them, download speed matters. The two directions can have very different capacities on the same broadband service.
For a UK connection, check the actual ISP, access network, product and location. FTTP alone does not establish the upload rate or guarantee a symmetrical retail service. The figures below are assumed transfer rates, not claims about available packages.
Techie Broadband calculation: initial upload versus daily changes
Assume an initial dataset of 100 GB and a subsequent day's transferred changes of 2 GB, both using decimal units. Transfer time equals bytes × 8 ÷ bits per second. These are payload-only floors at a constant rate, without overhead, compression, contention, scanning or service-side delays.
Assumed upload rate | Initial 100 GB | Subsequent 2 GB |
|---|---|---|
20 Mbps | 11 hours 6 minutes 40 seconds | 13 minutes 20 seconds |
100 Mbps | 2 hours 13 minutes 20 seconds | 2 minutes 40 seconds |
500 Mbps | 26 minutes 40 seconds | 32 seconds |
A long first upload therefore does not automatically mean the everyday workload is impractical. Equally, if the objective is to publish 2 GB within ten minutes, the minimum average payload rate is about 26.7 Mbps. A 20 Mbps upstream cannot meet that target even before overhead. Allow additional capacity for real conditions and other traffic.
Do not confuse bandwidth with achieved application throughput. A service can limit transfers, a client can impose a rate cap, and WiFi, storage, CPU or the network path can become the bottleneck. A “processing changes” status may represent local indexing rather than a stalled broadband upload.
One upload can lead to several downloads
In a simple cloud-backed example, one laptop uploads 2 GB of new content and four other devices each download a full 2 GB copy. The origin uploads 2 GB; the other devices download 8 GB in total. Across those access connections, 10 GB of payload moves, although the new logical dataset is only 2 GB.
This calculation assumes one upload, no peer-to-peer or LAN distribution, no compression and all four recipients fetching everything. On-demand access may download less. If recipients share one broadband connection, their simultaneous downloads share its capacity; four clients do not each receive the line's entire bandwidth independently.
Large numbers of small files create a different workload from one large file. Directory enumeration, metadata requests, antivirus scanning, file opens and service request handling can dominate. Microsoft's OneDrive limitations documentation identifies filename, path, item and account restrictions that can prevent or slow synchronisation. Check the current rules for your client and account rather than apply another product's limits.
An unrestricted upload can also fill queues and increase delay for calls or interactive traffic. This relationship is explained in the bufferbloat guide. If sync disrupts other activity, measure loaded latency and consider a suitable client upload cap, scheduling or correctly configured queue management. There is no universal cap percentage that works for every connection.
Is cloud sync suitable for a NAS or active database?
A supported NAS connector can synchronise selected folders while the NAS remains online, but a desktop client's folder rules do not automatically support network shares. Microsoft's current OneDrive documentation, for example, excludes using a network or mapped drive as its sync location. Check the actual connector instead of building around an unsupported shortcut.
Ordinary file sync is also not a substitute for database replication or an application-consistent backup. An active database, virtual machine image or multi-file project can change while files are being copied. A collection of individually transferred files may not form a consistent application state.
Use the application's supported export, snapshot, backup or replication mechanism when consistency requires it. If you synchronise the resulting closed export files, verify both their contents and restore procedure. File-level completion does not establish that a restored application will start correctly.
Avoid configuring two unrelated sync clients to manage the same working folders unless both explicitly support that arrangement. Apple specifically addresses using another service with iCloud Desktop and Documents and directs users to separate locations or disable the overlapping feature. Unclear ownership makes deletion and conflict diagnosis harder.
Security and account ownership
Protect the account with strong authentication, keep recovery methods usable and restrict sharing to the people who need it. A public sharing link, an organisation-owned folder and a private personal folder have different access and ownership implications. Check who retains the files when a colleague leaves or an account is closed.
Encryption in transit protects a network connection; it does not by itself mean the cloud provider cannot read stored contents. Check the service's encryption model and who controls the keys. If you add client-side encryption, test recovery and understand the effect on previews, sharing and duplicate detection rather than assume every feature will remain available.
Keep devices and sync software updated. Review connected devices, third-party access and local permissions. For a UK organisation, use an approved account and storage arrangement for customer or staff data, with clear access, retention and recovery responsibilities. Selecting a cloud region alone is not a complete data-handling policy.
How to check a cloud sync setup safely
Use a disposable test folder with harmless files before trusting important work. The objective is to prove the behaviours you depend on, including recovery, rather than simply observe a green icon.
Confirm the account, device, folder scope and one-way or two-way policy. Identify the local path and the corresponding cloud location.
Create a small test file. Wait for the source client to finish, then verify the contents through the cloud service and a second device.
Edit it on the second device and confirm that the expected version returns, if the configuration is meant to be two-way.
Mark a test file for offline retention, wait for its download and try opening it after disconnecting from the internet.
Test concurrent offline edits only on disposable copies. Identify how conflicts are reported and preserve both versions before resolving them.
Delete a disposable test file and observe the configured propagation. Restore it using the documented recovery mechanism and open it to check the contents.
Test the separate backup or export you intend to rely on. Confirm it contains real file contents, not merely online-only placeholders, and that an earlier version can be recovered if required.
Record the result for the actual application and file types you use. A successful text-file test does not certify a live database or every file in a large photo library. Revisit the checks after a major client, account or policy change.
Troubleshooting cloud sync that is slow, stuck or missing files
Start with the specific file and the specific direction. “Not syncing” could mean the upload failed, the wrong account is open, the destination has not downloaded it or the file is correctly excluded.
Wrong or missing version: compare local contents, cloud contents and the second device. Check account, folder scope, conflicts and whether the item is a shortcut or a full file.
Pending changes: read the client error and activity log. Check paused sync, authentication, quotas, permissions, unsupported names or paths, and files still being written by an application.
Offline opening fails: confirm that the full content was retained locally before disconnection. A thumbnail or folder listing is insufficient evidence.
Low transfer rate: distinguish scanning from transfer, inspect client caps and storage/CPU load, then test a suitable wired connection and the relevant upload or download direction.
Missing space or unexpected usage: distinguish local disk, cloud quota, cached content and retained versions. Freeing one does not necessarily free the others.
Unexpected deletions: stop making changes, pause affected sync activity where safe, preserve remaining copies and inspect recovery options before reconnecting or rebuilding clients.
Before resetting a client, clearing its cache, unlinking an account or changing storage modes, preserve any unsynchronised local work outside the affected tree. Google specifically warns that switching modes or deleting the previous mirrored folder before uploads finish can lose unsynced changes. Avoid repeatedly restarting a broken relationship without first identifying which location contains the data you need.
Frequently asked questions
Does cloud sync work without the internet?
You can often work offline on files whose full contents are available locally, if the application supports it. Changes wait to synchronise after reconnection. Online-only placeholders cannot supply missing file contents without a connection.
Does deleting a synced file delete it everywhere?
It can. Two-way services commonly propagate deletions within their configured scope, while one-way tasks follow their own deletion policies. Use the service's local-space control instead of ordinary deletion when you only want to remove a downloaded copy.
Is cloud sync enough as a backup?
It can help after a device failure, but synchronisation alone may also spread unwanted edits, deletions or corruption. Check version retention and maintain recovery copies protected from the same failures, then test that you can restore them.
Why is a cloud file visible but unavailable offline?
The visible item may be an online-only placeholder containing metadata rather than the full file. Explicitly make the file available offline, wait for completion and test opening it before relying on it without internet access.
Does cloud sync transfer the whole file after every edit?
It depends on the client, service and file format. Some implementations reuse unchanged blocks; others transfer a complete modified file. Measure the actual transferred data and consult the relevant product specification.
Will faster broadband make cloud sync instant?
No. It can shorten a transfer that is limited by the relevant network direction, but scanning, service limits, storage, conflicts, offline devices and download policies still affect when data becomes usable elsewhere.
The key takeaway
Cloud sync is most useful when you know what is synchronised, where full copies exist and how unwanted changes can be recovered. Check those behaviours with disposable files, size the connection for the data that actually moves and keep recovery separate from the assumption that every device showing the same folder must be safe.

