> From: zfs-discuss-boun...@opensolaris.org [mailto:zfs-discuss-
> boun...@opensolaris.org] On Behalf Of Sašo Kiselkov
> This is so profoundly wrong that it leads me to suspect you never took
> courses on cryptography and/or information theory. 

More inflammatory commentary, and personal comments?  Saso, seriously, knock
it off.

> The size of your
> storage pool DOESN'T MATTER ONE BIT to the size of the key space. 

Are we still talking about SHA256 collisions in a zpool?  If so, the size of
the storage pool (at least, the used blocks inside the pool) definitely has
an effect on the probability of a collision.

The first block written to pool, gets a checksum.  There is a
zero-probability of collision, because it's the first one.
The second block written gets another checksum.  It has a 2^-256 probability
of collision with the first one.
The 3rd block has a 2*2^-256 probability...
The 4th block has 3*2^-256 probability...

This is the birthday problem.  If you don't know it, look it up on
wikipedia.  Each block written has a higher (but still very low) probability
of collision with any other block that was previously written, until ...

When you write your approximately 2^128th block, you have a 50/50 chance of
collision with any other block.  Fortunately, this amount of data is far
greater than all the storage in the world combined, so this probability can
never be reached with earthly means.

For all Earthly data sets, assuming no SHA256 weaknesses are being
exploited, the probability of a collision is still extremely small, but the
more data blocks in the data set, the higher the probability is.

zfs-discuss mailing list

Reply via email to