Watch seven files land on an empty 80-block disk as clean unbroken runs, watch three of them get deleted, and watch the next big write be forced to scatter into four separate gaps — then press Defragment.
An 80-block disk drawn as a grid, 4 KB per block, and a scripted thirteen-step history played over it. Each colour is one file; a free block is an empty outline.
Press Play the history and watch, or press Step to walk it one event at a time. Six files write first, and while the disk is still empty every one of them lands as a single unbroken run — that is what an unfragmented disk looks like, and it is the baseline everything after it is measured against. Then three files are deleted. Nothing moves to close the holes they leave, because closing a hole would mean rewriting everything after it.
The tenth step is the one to watch. project.mov needs 34 blocks. There are 47 free, but the largest single run is only 20, so the file system fills the first gap, jumps to the next, fills that, and jumps again. The file ends up in four pieces — and because each write fills one block at a time, you can see the jump happen rather than only see the result. Fragmentation is not something that went wrong with the disk; it is what deleting things does to free space.
Click a file in the table to isolate its blocks and keep it isolated. That is the clearest way to see that all four orange clusters are the same file.
The numbers are a stated model, not a measurement, and the model is printed under the readout so you can argue with it: reading a file costs one seek per piece at 10 ms each — a hard disk head move plus a platter rotation — plus 0.05 ms to stream each 4 KB block. EXTRA SEEKS is the sum of pieces minus one across every file, which is the part that is fragmentation’s fault and nothing else’s.
Then press Defragment. Every file is rewritten into one contiguous run, extra seeks fall to zero, and the modelled read of the whole disk drops by about 43%. It does not drop to zero, and should not: a contiguous file still costs one seek to find.
Two honest caveats, both stated on screen. Allocation here is first-fit, the simplest policy that fragments for the real reason; NTFS and ext4 both try considerably harder to keep files together, which is much of why modern desktops rarely need this done by hand. And every cost in the model is a physical head moving, so every conclusion is about hard disks. An SSD has no head and no platter, reaches any block as fast as any other, and gains nothing from defragmentation — running a defragmenter against one just spends write cycles.
Nothing here is graded and nothing is submitted. Press Reset to start the history over.
TechKnowSurge builds IT and cybersecurity professionals through hands-on, concept-first training built around real understanding — not memorization. Free interactive tools, structured programs, and 25+ years of real-world experience, all in one place.
Explore free tools and programs →