Talk:Replication (computer science)

From Wikipedia, the free encyclopedia

Contents

[edit] CORBA

What about open standards like FT CORBA from OMG? —Preceding unsigned comment added by 195.72.96.178 (talk)

[edit] External links

What about external links like:

—Preceding unsigned comment added by Kurtseifried (talk • contribs)

I would prefer as few external links as possible. Wikipedia is not a link directory. External links sections often get out of hand, especially when one provides such an obvious space for advertising. -- intgr 08:28, 25 January 2007 (UTC)

[edit] Active/active replication POV/advert

I have tagged the "Active/active replication" section with {{advert}}, since it currently claims that WANdisco has exclusively invented non-master/slave replication, and seems to imply that theirs is the only possible/existing approach, and that it has no drawbacks. However, no reliable sources supporting its uniqueness are presented, and I fail to take this for granted. Claiming that it's a "new mathematical theory" further trips my buzzword meter.

This is coupled with the fact that there's also a Wandisco article with these same claims, which was tagged advert for a second, but subsequently had the tag removed by an anonymous user (diff).

Some factual information about their implementation is certainly needed to make this entry encyclopedic. Reflecting the advantages/disadvantages would be a start. The obvious question arises that how they resolve conflicting updates (e.g., things that would otherwise cause lock waits or deadlocks). -- intgr 10:24, 24 January 2007 (UTC)

I have added clarity that WANdisco's technology works efficiently over a WAN. —The preceding unsigned comment was added by Owlsfan (talk • contribs) 20:54, 8 February 2007 (UTC).—Preceding unsigned comment added by Owlsfan (talk • contribs)
I can address this a bit. I am a co-Founder/Inventor of WANdisco and was the CTO till the last month (Jan'07). WANdisco implements conflict avoidance using its DCone (Distributed Coordination engine) as mentioned in the WANdisco whitepaper and the pending patent.

--Brahul1 07:57, 19 February 2007 (UTC)

[edit] disk storage replication

on Semi-synchronous replication: considering that synchronous doesn't always refer to a temporal synchronicity but rather a structural one, my guess is that "semi-synchronous" simply refers to a unidirectional replication when thats sufficient for the use case, that is if you consider one replicant as a source and (an)other(s) as target(s). common cases are copycat behavior and any kind of oneway communication, where a replicant (target) needs only to consume external information rather than to advertise its own. on the other hand, unidirectional replication comes in handy, when the source replicant is unable to consume information (write protection). Robin479 (talk) 14:39, 29 February 2008 (UTC)

Hmmm, I don't agree. This isn't mentioned in the article, but currently popular disk storage replication methods are predominantly unidirectional. So all three designations—synchronous, semi-synchronous, asynchronous—usually refer to unidirectional transfers. --Kubanczyk (talk) 20:14, 29 February 2008 (UTC)