Re: LDAP/x.500 Search Benefits??

Jochen Keutel (keutel@u9ytb.iz-darmstadt.telekom.de)
Mon, 23 Sep 1996 10:20:36 +0200 (MESZ)

Hello,

> I am trying to judge the cost/benefits as compared with RDBMS's.
> Everyone says things like LDAP is optimized for read many write
> infrequently. How does this optimization manifest itself?
>
> How is data stored?
> How are searches accomplished?

this depends on the implementation of the LDAP/X.500 server. Storage of
data isn't part of the standards.
A lot of X.500 products are Quipu-based - they store data in normal
UNIX files and load them into memory when getting started. Other X.500
products use RDBMS as backend (the one I use uses C-ISAM).
Umich's slapd can use ndbm, gdbm, Berkeley DB as backend.

> I have a database with > 20 Million records representing individuals.
>
> In general would a properly conceived RDBMS setup outperform an x.500 or
> LDAP setup for a white pages application?
>
> If so, would the difference be substantial?

Again - it depends on the server. I'm currently experiencing with 1 million
entries in one X.500 server - certain searches are as fast as the
corrosponding SQL query in the original (Informix) database.

> I understand how RDBMS are optimized for certain searches by building
> indexes on selected fields (i.e. name or x.509 certificate) but how does
> a x.500 or LDAP server optimize for the common searches?

As far as I know Quipu it's optimized for searches for name parts
(means searching for attributes which are part of the DN). Searching for
other attributes could be a performance decrease.
In my C-ISAM-based X.500 server you can put indizes on attributes of
your choice. The same is possible with slapd (and, for example, gdbm).

> Can LDAP be substantially improved by having more servers? more memory?
> more processor? Is it I/O bound?

Again - implementation dependent. If using a database it's of course I/O
bound - so more memory / processor should help.
More servers: If you distribute your data on several servers then the
mechanism of referrals / chaining should increase performance - but I haven't
seen any documented measurement on this. ( I will do some measurements
when I find time for that.)

> Has anyone seen any documents which have addressed the size limits of
> x.500 or LDAP implementations as compared to relational databases?
>
> Finally, if anyone has done a commercial comparison of large database
> implementations using LDAP/x.500 vs. RDBMS technology I would be
> interested to hear about it as well.

Me too. My general impression is that current X.500 implementations are not
optimized for performance - a well-designed RDBMS will be faster than
a X.500/LDAP server.
I feel that performance is an important topic - most companies do have a RDBMS
containing data of their employees, and you can't convince them to use
X.500/LDAP if there are significant performance disadvantages - even if
X.500/LDAP has other advantages.

>
> Thanks for any input, I apologize for the barrage.
>
> Neil
>
>
> ______________________________________
> Neil Spiegel
> Atlas Consulting Inc -- Toronto, Ontario Canada
> V: 416-534-9792 F: 416-534-7888
> nspiegel@sympatico.ca

Bye,
Jochen.
------------

Dr. Jochen Keutel currently at: Deutsche Telekom
duerr com-soft IZ Darmstadt

Phone: +49 6151 818 579

e-mail: keutel@u9ytb.iz-darmstadt.telekom.de
100.37805@germanynet.de
X.400 : /C=de/A=dbp/P=telekom400/O=dmst03/OU1=08/S=osys-02
WWW : http://www.geocities.com/WallStreet/3454