Hi all,

Given the discussion about the use of bits in the URI for CoMI,
it is reasonable to look at the probability that a hash clash occurs,

In the attached table I set out the clash probability of the number of names in a server versus the hash size.
Currently the hash size is 30 bits and is proposed to be reduced to 29.

The table show that by removing a bit in the hash, the clash probability increases with a factor 2

with 5.000 names and a 30 bit hash, the clash probability in one server is 10% With 30.000 names and a 32 bit hash value the clash probability in one server is also 10%

A higher probability than 10% looks unwanted to me and a probability of 1% looks adequate to keep the client re-hash table size small.

To estimate the size of the hash, the number of names per server should be estimated. A hash size of 29 seems reasonable when the expected number of names per server remains smaller than 3000.

Has anybody an idea about the number of names in a reduced resource server?

Looking forward to your reaction,

Peter





--
Peter van der Stok
vanderstok consultancy
mailto: [email protected]
www: www.vanderstok.org
tel NL: +31(0)492474673     F: +33(0)966015248

Attachment: hash clashes.docx
Description: MS-Word 2007 document

_______________________________________________
6tisch mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6tisch

Reply via email to