As posted on Ed's blog, there's a lot of talk going on right now about DAOS. I thought I'd try to clarify some of the benefits and potential pitfalls as clearly as possible to help some of you make the decision.

First the Potential Benefits

Most sites are reporting between 40 and 60 percent reduction in disk storage requirements without making any policy changes or requiring users to change behavior. I personally believe you will be hard pressed to find any site with more than 50 mail users that isn't going to see similar savings unless that site already has some pretty stringent rules in place or unusual work patterns.

Saving 50% of disk space can mean a proportional savings in disk I/O -- the number one determining factor in server performance.

Design partners have been heavily testing DAOS for quite a while, and people I know and respect a great deal have really tortured it by crashing servers and even purposely damaging files to see how robust it is. It's proving to be very reliable under real world conditions.

DAOS is the future. IBM could have done it behind the scenes and hidden it from you -- just called it the newest ODS (On Disk Structure) version. That's not the way IBM does things, but don't mistake the fact that DAOS is going to be key to many additional space and performance saving features in the future. As a Design Partner, I've been privy to some of those things and am only allowed to say "wow".

Now the Potential Pitfalls

Gone for good are the days when you could just copy an NSF file at the operating system from one machine to another. You shouldn't be doing that now, but you surely can't do it with databases using DAOS

Backup and Restore is more complex. There is no getting around this point. For now, best practices and tested solutions haven't really emerged around this issue. There are tools which support it -- Tivoli and Symantec both have backup products that fully support Domino 8.5. I'm just not sure enough time has passed that we know all the issues. For mission critical databases -- at least for now -- I'm replicating to a non DAOS server and taking backups from that.

Those are the issues as I see them. A few people are out there claiming that "well run" environments won't see savings because the "pruning" process is batched rather than done in real time. I completely disagree. I think its a fallacious argument based on totally incorrect assumptions. We'll see if someone can prove me wrong -- but I'm not holding my breath.