What is the relative complexity of reads and writes in OpenLDAP? Are modifications as expensive as creations? I'm aware that writes are not as cheap as reads, and anecdotally I've seen figures like ten-to-one -- the usual justification being that writes force the lookup tree to be rebalanced. Which lookup tree? The articles I've read (their URL escapes me at the moment) don't really say, but I assume it's the DN lookup tree.

I'd like to know: o How does the complexity of creation vary? Maybe the tree is rebalanced only once in awhile. o Do attribute modifications have the same complexity as creation? Probably not, if the issue is rebalancing the DN lookup tree.

I will be able to ask more intelligent questions in the coming month, when I put together benchmarks and actually learn the OpenLDAP API -- so I have not asked for any ratios, for example (comparing number of `ldap_search_ext()` requests to number of `ldap_mod_ext()` requests is disingenuous, since the latter gets a DN to start with and the former does not.)

At this point, I'm working out a proposal to propose another proposal for using LDAP for horizontally scalable storage of small records -- I need to be able to say something about write performance, but I'm not able to put work into the benchmarking yet.