From: Eric B. <er...@eb...> - 2007-03-31 20:47:07
|
Hi, > I would like to start thinking about doing a feature freeze for version > 2.2.0. There are three major aspects of this: > > 1. Completing the work on features already implemented. > 2. Deciding what additional features to implement. > 3. Determining a rough timeframe. > > For item 1: The work that I see that remains to be done on existing > features is: > Concerning batch insert: > - Getting the documentation that Marc is working on to me. > - Doing some really heavy stress testing of the code. > - Deciding if we want to provide a way to dynamically disable the new code > (probably not if the stress testing goes well). > > Concerning other features: > - I think they are all pretty well covered, some that I have implemented > still need documentation. > > Do you or anyone else have any additional items to add to this? No, it's looking good, i have wrote the checkfilechanges doc today. > ==== > > For item 2: I have a couple of features in mind that we might want to > implement for version 2.2.0. I've listed them in the order of priority I > see: > > a. Copy (similar to Migration, but not purging the original copy) good > b. Despooling attributes at the same time that we despool data to the > Volume. This requires creating a second thread for the job -- something > totally new and not without a lot of possible race problems. Yes, it will be useful. > c. If item 2.b is implemented, I will also want to implement commit of the > attributes when the partial despooling of the attributes is done. Did you mean "when the partial despooling of the data is done" ? Yes it will be very cool. > d. Implement spooling data from the FD while despooling an existing file > from the same FD. This is a lot more tricky than the project above to > despool attributes while despooling the data. But it's more useful :), we should simplify this i think, i have ideas about this, but i have to work a little. > By the way, I am pretty sure that we can implement all the new features > above in a reasonable timeframe, with the possible exception of spooling > data from a FD while simultaneously despooling data (and attributes). > > The above with all the work we have already done will make a really nice > packet of performance enhancements :-) > > Depending on what we decide under item 2 we could release 2.2.0 in one > month (no new features) to three months (all four new features). I vote for wait 3 months :) > In addition to the above, 2.2.0 will include all currently implemented > features -- some of the more important being: red/black binary tree code > for restore, database name/user configurable, Bacula Admin Tool -- bat > (first cut), enhanced pruning algorithm, Scratch pool recycling > enhancements ... > > I address this to Eric because most the remaining work on existing features > fall on his plate, and he has implemente the most important item mentioned > above (batch insert) as well as having proposed implementing item 2.d. > That said, I am interested in feedback from everyone. > > As for other projects in the works (and there are some), I have no > tentative schedules yet, so they are not included, but that doesn't mean > they are excluded. I think that my "relocation file" code will be ready soon. Bye |