Unit standardisation

[eileen-kuehn]

We should decide the standardisation of units through the simulation. We should further decide on the i, so either KiB or KB (no matter what if K or M or G or ...). Currently I think we use GB.

Concerning the storage size, I recommend deciding on one data size unit and sticking to it. If we use Bytes then float is enough to represent up to 8 PiB precisely increasing with the prefix (e.g. 8 EiB for KiB base); Since there are no fractional Bytes, using int would allow for arbitrarily large and precise storage size even when using Bytes.

Originally posted by @maxfischer2781 in #53

[maxfischer2781]

Using plain bytes has the least problems in standardisation (we do not need to decide on KiB versus KB etc.) and is always precise when using int. However, it might degrade to float during naive computations, which is imprecise beyond 8PB.

[maxfischer2781]

After some testing, I strongly favour int in Bytes. The only issue is silent coercion, but that can be covered by type checkers easily.

In contrast, if we use float in GB we have no way whatsoever to check against accidentally using a different unit.