Review build Not the live site — unlaunched products, unfinished copy, no prices. comsky.ai →
Comsky Backup
Sign in Try Backup free
What gets protected

Not a sync folder. A versioned history of every drive.

File sync gives you the latest copy of a file — including the latest encrypted copy, if something got to it first. Comsky Backup keeps every restore point separately, so there is always a version from before.

RESTORE POINTSACC-SRV-01 · 61 KEPT
02:00 today newest verified
02:00 yesterday verified
02:00 · 14 Sep 3 files skipped
02:00 · 13 Sep no shadow copy
02:00 · 12 Aug oldest kept verified
Every run adds a restore point. Retention decides which ones age out — nothing overwrites. A point is only removed once it is outside both the age and the count.

A whole drive is a valid source.

Point a policy at C:\ and D:\ and the agent walks the tree and takes what it finds — minus a built-in exclusion list of Windows scratch files. The accounts data, the shared folders, the user profiles and the application directory nobody documented are in scope without anyone having to remember them. Or name specific folders instead, with your own exclusions on top.

Open and locked files are read through a shadow copy.

A backup does not have to wait for everyone to go home. Each run asks Windows for a shadow copy of the volume and reads the files out of that, so a database or a mailbox that is mid-write is read as it stood at one instant rather than being skipped. If a run finds a shadow already stuck from a crashed one, it clears it and carries on.

Schedules and retention you actually control.

A policy sets the frequency — hourly inside a window, daily, weekly or monthly — the compression, the exclusions, and how long restore points survive, in days and in number. It then applies to every device attached to it. Change the policy and the whole group follows. No per-machine scheduled task to maintain, and no cron job that quietly stopped firing in March.

Only what changed leaves the machine.

Repeated content inside a machine’s own history is stored once, and after the first run only new and changed content is uploaded — a nightly backup of a server that changed by 2 GB moves roughly 2 GB, not the whole drive again. The console shows the real split for every run: scanned, after dedup, after compression, uploaded. It is a measurement, not a vendor ratio.

Excluded by default, before you configure anything:

pagefile.syshiberfil.sysRecycle Binbrowser caches*.tmpnode_modules
The honest limit

When the shadow-copy service will not start.

A backup product that only describes the good path is describing a demo. Here is what happens on the machine where Windows will not give the agent a shadow copy at all.

The run still goes ahead without one. Files that were locked at that moment are skipped, listed by name against the job, and the run is flagged rather than filed as a clean success. The restore point is badged in the picker, so when you come to restore from it you can see which runs excluded locked files and how many files a run could not read.

A stuck shadow is cleared

If a run finds a shadow copy already stuck from a crashed earlier run, the agent clears it and carries on rather than failing and waiting for a person.

Skipped files are named

Not a count in a log somewhere. The job carries the list, so the question of what did not make it in has an answer you can read.

A flagged run is not a green tick

A run that had to skip files is flagged. A run that stored materially less than the pre-scan expected is flagged too. Both are visible in the console without opening anything.

Policies

Set it once, and the whole group follows.

A policy is where the frequency, the sources, the compression, the exclusions and the retention live. Attach devices to it and they inherit all of it — and two policies can cover the same folder on the same machine with different retention.

NIGHTLYACTIVE
runsEvery day at 02:00
sourcesC:\ and D:\ — whole drives
keep90 days · max 120 points
devices9 servers
Shadow copyCompression: high
LAPTOPSACTIVE
runsHourly, 08:00–20:00, weekdays
sourcesC:\Users — profiles and Documents
keep30 days · max 200 points
devices24 laptops
Shadow copyCheckpoint 10m
SHARED DRIVESACTIVE
runsEvery day at 23:00
sourcesE:\Shared
keep365 days · max 400 points
devices5 file servers
Long retentionCompression: maximum
RETAIL1 DEVICE OFFLINE
runsEvery day at 22:30
sourcesC:\POS — excludes *.log
keep14 days · max 30 points
devices3 terminals
Compression: normal

Two policies can cover the same folder on the same machine with different retention — hourly points kept for a fortnight, and a nightly one from that same folder kept for a year. They share one deduplicated repository, so the second schedule costs very little to keep.

What a run actually moves

Scanned, deduplicated, compressed, uploaded.

Four numbers per run, measured from the job rather than quoted from a brochure. This is what a nightly run on a file server looks like after the first one has been and gone.

NIGHTLY · ACC-SRV-01 · 02:00412,900 FILES
scanned on disk1.9 TB
new and changed3.4 GB
after deduplication2.6 GB
after compression2.1 GB
uploaded2.1 GB
STORAGE IS BILLED FROM THE OBJECT STORE, AFTER DEDUPLICATION AND COMPRESSION · NOT FROM WHAT WAS SITTING ON DISK

Deduplicated per device

Repeated content inside a machine’s own history is stored once. The second policy pointed at the same folder shares that repository, which is why a second schedule costs very little to keep.

Interrupted, not failed

A run cut short by a dropped connection, a sleeping machine or the agent updating itself is recorded as interrupted and resumes from its last checkpoint instead of starting the upload again.

A measurement, not a ratio

The console shows the real split for every run. Nobody has to trust a figure from a data sheet about how much a backup product will compress your particular files.

Questions

About what does and does not go in.

Can I back up a whole drive, or only named folders?

Either. A backup source can be a whole volume or a folder you name. Pointed at a volume, the agent walks the tree and takes what it finds, minus a built-in exclusion list of Windows scratch files — the page file, the hibernation file, the recycle bin, browser caches, *.tmp and node_modules. You can layer your own exclusions on top of that per policy.

What happens to files that are open when the backup runs?

Each run asks Windows for a volume shadow copy and reads out of that, so a database or mailbox that is mid-write is captured as it stood at one instant rather than skipped. If a crashed earlier run left a shadow stuck, the agent clears it and carries on. If the shadow-copy service cannot start on that machine at all, the run still goes ahead without one — locked files are skipped, listed by name against the job, and the run is flagged rather than filed as a clean success.

How long are restore points kept?

As long as the policy says, expressed both as an age in days and as a count of points. A point is only removed once it is outside both. The expiry itself runs on the server with a credential the endpoint never sees, so nothing on the machine decides when its own history ends.

Can the same folder be in two policies?

Yes, and it is a reasonable thing to do. Hourly points kept for a fortnight and a nightly point from that same folder kept for a year is a common shape. They share one deduplicated repository, so the second schedule costs very little to keep.

Is this a disk image?

No. It is file-level backup — including every file on a drive, if that is what you point it at. There is no volume image, no recovery media and no bare-metal restore. If a one-click whole-machine rebuild is the requirement, you need an imaging product instead.

Every device name, policy and figure on this page is demo data.

Start with Comsky Backup.
Grow into the ecosystem.

One account opens every Comsky product — and one invoice covers them, however many you run.