Convert TB to kB

A terabyte is a trillion bytes. A drive sold as 1 TB is exactly that — the space that seems to go missing is a unit disagreement, not a defect.

A kilobyte is 1,000 bytes when a standards body defines it and 1,024 when an operating system does. This page uses 1,000; the kibibyte is the unit that means 1,024.

Going from terabytes to kilobytes moves the decimal point 9 places to the right and changes nothing else. There is no factor to remember and no rounding to decide: the digits stay in the same order and only their position changes.

1,234 TB is 1234000000000 kB — same digits, moved along. That is worth knowing because it is the one kind of conversion you can check at a glance: if the digits of the answer are not the digits you started with, something other than the conversion has happened to them.

Decimal makes the ratio exact and unmemorable: a kilobyte is 10³ bytes, a terabyte is 10¹², and there are 1,000,000,000 kilobytes in a terabyte. That is nine places of the decimal point in a single step, which is more than most people can carry reliably in their head, and it is the reason this particular pair is worth typing into a box rather than estimating.

The scale is also why the two units rarely appear in the same sentence outside a report. Terabytes describe volumes, arrays and purchases; kilobytes describe log lines, packets, and configuration files. Nothing a person handles is naturally measured in both, and no intuition connects them: a billion of anything is a count nobody can picture, which is why the answer here has to be calculated rather than estimated. When both turn up together, it is almost always because a tool chose to report a large thing in a small unit, and the interesting question is which small unit it meant.

df and du report in 1K blocks by default, and those blocks are 1,024 bytes. The convention is older than the kibibyte, which was only named in 1998, and it has never been changed because too much tooling parses the output. So a one-terabyte filesystem appears as 976,562,500 blocks, not 1,000,000,000 — and 976,562,500 × 1,024 is exactly 1,000,000,000,000, which is a satisfying way to confirm the arithmetic on a real machine.

Both accept flags that change the unit, and the safest habit is to ask for bytes and convert once yourself. Where that is not possible, record the block size next to any figure the moment you copy it out, because nothing about the number afterwards will say which convention produced it. A total that has passed through a spreadsheet and an email has lost that information permanently.

The kilobyte and the kibibyte differ by 2.4 per cent, the narrowest gap in the data units, and at kilobyte scale it is genuinely ignorable. Multiplied by a terabyte it stops being ignorable: 1,000,000,000 decimal kilobytes against 976,562,500 kibibytes is a difference of 23,437,500 units, which is 24,000,000,000 bytes — 24 GB, or a fifth of a modern game.

That is what makes this pair different from every other one in the decimal data set. Elsewhere the advice is that 2.4 per cent is noise beside compression ratios and block rounding. Here the multiplier is nine orders of magnitude, and a capacity plan built on the wrong kilobyte is out by more than most people would tolerate in a hardware budget.

The other reason to make this conversion is to turn a capacity into a count. A terabyte is a billion kilobytes, so it holds 250 million 4 kB records, 500 million 2 kB log lines, or a billion 1 kB rows, packed end to end. Those figures are useful precisely because they are countable: a system producing fifty million events a day at 2 kB each fills a terabyte in ten days.

The packing assumption is the one to check. Records inside a database file or a log file are contiguous and the arithmetic holds. Records that are separate files are not: on a filesystem with 4,096-byte blocks, every file under 4 kB occupies a whole block, so a billion 1 kB files need about 4 TB rather than 1. Whether the small things are rows or files changes the answer by a factor of four.

A quota report is not a sum of file sizes. Quota accounting tracks blocks allocated, because that is what the filesystem charges against a volume, and the figure includes the wasted tail of every partly filled block. A user with a million small files can be over a quota that their file sizes add up to comfortably under, and there is no error anywhere in that.

Converting one therefore needs the block size as well as the number. Ten million blocks is 10.24 GB at 1,024 bytes, 40.96 GB at 4,096, and 5.12 GB if the system happens to report 512-byte units — a spread of eight to one, decided by a parameter that appears nowhere in the number itself. Ask the filesystem what its block size is before doing anything with a quota figure.

The precision itself is not at risk. A billion, and every figure on this page, is an integer far inside the range that spreadsheets and standard floating-point arithmetic represent exactly, so nothing is lost in the calculation. What goes wrong is display: a general-format cell renders 1,000,000,000 as 1E+09, a narrow column shows a row of hashes, and a number retyped from either of those has silently become an approximation.

Format the column as a plain integer before the numbers arrive, keep thousands separators so that a missing or extra zero is visible, and label the unit in the header rather than in a note somewhere else. Nine-digit figures with no separators are unreadable, and a capacity plan that is out by a factor of ten because a zero was dropped is a far more common failure than any unit confusion on this page.

Kilobytes are a poor intermediate unit at terabyte scale, and the moment more than one figure is involved they should be abandoned. Byte counts are integers, unambiguous across every system and tool, and exactly comparable; kilobyte figures carry a divisor that has to travel with them and usually does not. Sum in bytes, compare in bytes, and convert once at the end for whoever has to read the result.

The exception is the case this page was written for: a report that is already in kilobyte blocks and a requirement that has to be checked against it. There, converting the requirement down to match the report is quicker and less error-prone than converting a million rows of report up. Do the conversion in the direction that touches the fewest numbers, and note the block size beside the answer.