Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 31 Dec 1998 12:04:47 -0800
From:      dboreham@netscape.com (David Boreham)
To:        freebsd-current@FreeBSD.ORG
Subject:   Re: NOW/MOSIX/Beowulf
Message-ID:  <368BD8DF.C478B885@netscape.com>
References:  <368BCBE7.BCC1FFB5@netscape.com> <19981231142400.X4306@ns1.adsu.bellsouth.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Christian Kuhtz wrote:

> > The current Netscape LDAP server supports multi-server
> > replication where each distinct DIT subtree must be
> > mastered on one and only one server.
> 
> But you don't support multimaster.  Which is a crucial in almost all of our
> applications.  Replicating read-only tries is only good if the directory is
> used as a reference, but not when the intention is to make it an active
> container of information.

Multimaster replication will be in the next major release.

Note that MMR does not solve all the problems associated
with distributed databases. For example, a client performs
an update on one master. It then queries the directory service
to read back the modified entry. Due to load balancing
it reads from a different server which has not yet received
the update. Client therefore sees inconsistent data.

In the end there are fundamental laws of physics
involved, which no fancy protocol and no army of
wizzard coders can transgress.

> > Thus read load can be distributed across multiple
> > replication consumers easily.
> 
> But no commits/writes.  All commits on a Netscape DS have to be made on a
> 'central' server (the owner/master of the tree).  I think that's what the
> previous poster might've meant.

Not quite true. For a given subtree, all writes must
be made on a particular designated server.
Provided the data is partitioned across the tree,
multiple masters are a reality.
 
> That's a theoretical possibility..

<comments snipped>

Yes, I agree. 
A client should not need knowledge of the service topology.

Various extensions to the LDAP protocol, the X.500 information
model, and our server implementation have been considered to
"solve" this problem. Generically it's termed "the distribution
problem" in Directory circles. Prototype implementations
are in use on my.netscape.com and www.netscape.com
and webmail.netscape.com

Current product supports LDAPv3 referrals.
Clients may connect to an LDAP server of their
choice (perhaps determined by DHCP or DNS)
and submit a query. That server may refer the
client to one or more other servers, which
the client may contact in order to resolve the
query. Network traffic could potentially be
optimized by forwarding the query from
server to server rather than having the
client re-submit to referenced servers.
This forwarding is termed "chaining"
in Directory parlance, and could be implemented
by any LDAP server transparently to the client.
We have an in-house custom server which 
does chaining, but the current product does not.

In any event, the mechanisms for client
query interaction in a distributed environment
are present in LDAPv3. However, in order to
make distribution practical, some additions
are needed on the server side.
Currently referrals are statically configured.
Referral decisions should be made by the
server based on a real understanding 
by it of the service topology and partitioning
schemes employed.
Clients also need a way to ask a server for
the distinguished name of a new entry
they wish to add to the service.
These are areas in which work is ongoing.

Anyway, fascinating as this is, it's going well off topic.
Follow-ups should goto to ietf-ldapext@netscape.com

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?368BD8DF.C478B885>