> I  would think that DNS would be more efficient anyway, and with a 1
> million  address/day  dictionary  attack on a single domain, a small
> number  could  make  a  huge difference.

Our  best-equipped LDAP boxes can handle 1 million+ queries/_hour_, so
it's not a question of the protocol not being the best fit. However, I
do appreciate that those with DNS-heavy expertise may not wish to deal
with  the LDAP learning curve plus the addition of separate LDAP boxes
(or running additional daemons on their dedicated DNS systems).

> Anyway,  another  piece  of  complexity that makes the ORF recipient
> blacklist  somewhat  impractical  is  the  growth  in  the number of
> addresses listed. The interface is dog slow with 10,000 addresses in
> it due to what I suspect is renumeration when you use it to save the
> config  and  restart,  though  the  service  doesn't lag with simple
> restarts  nearly so much. I wonder about how efficient this might be
> compared to DNS checks on 127.0.0.1.

You  have  two  different  areas  here:  update  performance and query
performance.  The  fact that the recipient blacklist updates require a
service  restart  at all is why I consider this suboptimal relative to
LDAP,  which  is  updated  while  the  daemon  is  running. Given that
architecture,  though, it's perfectly understandable that there should
be  a  delay  as  the file is loaded into memory, and possibly indexed
from  there. Does the cache size impact performance unacceptably after
it  is  loaded?  Is  this  only  an  update issue (and therefore, more
significantly,  an uptime issue)? Local UDP lookups against an indexed
db  will be faster than in-memory lookups against an large _unindexed_
db,  but  are  the  in-memory  lookups  indeed  unacceptably slow once
loaded,  indicating  poor  or  nonexistent  indexing?  I know that the
answers  don't  change  your need for a lookup protocol that keeps the
service up and running, but they are helpful.

My  preference  for  LDAP  is bolstered, in any case, by the fact that
IMail  and OpenLDAP are already synchronized, providing a ready system
for outside queries. The remaining problem is one of client lookup and
cache implementation, ORF being the client under review. ORF uses what
could  be  referred  to  as a L2 cache (L1 being the LDAP daemon's own
memory cache in this lexicon), but I'm not sure how well this cache is
indexed,  and thus how it performs under load. Caching implementations
need  to  flex  with  database  size  and query patterns, or they will
actually  slow  access, and theirs isn't user-tweakable (though it may
indeed  be self-tuning; I'd have to get into some benchmarks to tell).
The LDAP clients we've built, to the contrary, may be deployed with at
least three caching scenarios:

-  have  no "L2" cache within the application, but rather use a local,
high-speed LDAP replica as a cache for a remote master

-  use  an indexed L2 cache of singleton lookups, either instead of or
in addition to a local replica

-  use  an  L2  cache  of full-domain lookups, either instead of or in
addition to a local replica

-  use  a  combination  of  load-balancing across remote servers and a
local L2 cache

In  order  to  have  such  options,  there's  also the question of how
lookups  are  performed:  ORF  isn't optimized for singleton recipient
queries.  This is understandable given their total reliance on a local
L2  cache  of a general LDAP search that grabs all recipients at once,
but  with  some traffic patterns, you'd actually be better off with no
L2 cache at all, a local OpenLDAP replica, and singleton lookups.

In  sum,  then,  I must confess that ORF _may_ not be the best vehicle
for  the implementation of LDAP recipient lookups on a large userbase,
but the jury is still out. I'll try to do some testing next year.

--Sandy


------------------------------------
Sanford Whiteman, Chief Technologist
Broadleaf Systems, a division of
Cypress Integrated Systems, Inc.
e-mail: [EMAIL PROTECTED]

SpamAssassin plugs into Declude!
  http://www.mailmage.com/products/software/freeutils/SPAMC32/download/release/

Defuse Dictionary Attacks: Turn Exchange or IMail mailboxes into IMail Aliases!
  
http://www.mailmage.com/products/software/freeutils/exchange2aliases/download/release/
  
http://www.mailmage.com/products/software/freeutils/ldap2aliases/download/release/

---
[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail".  The archives can be found
at http://www.mail-archive.com.

Reply via email to