On Tue, Jun 7, 2011 at 6:35 PM, Christian Grobmeier <grobme...@gmail.com> wrote: >> 300000 downloads per day or per month? >> >> 52TB per month is still a lot... > > per day. > Look at this chart: > http://marketing.openoffice.org/marketing_bouncer.html
TL;DR: these bandwidth numbers are not actually that scary or important -- the OOo infrastructure does not serve up all that traffic, the many mirrors do, and traffic from the OOo infrastructure to the mirrors is very efficient since it efficiently pushes out new releases (with rsync). The big work in mirroring is setting up the process and shepherding the mirroring community. OOo has 118 mirrors. Apache has 294 mirrors. There is significant overlap -- a lot of sites that mirror apache downloads already mirror OOo downloads and vice versa. http://download.services.openoffice.org/mirrors/all.html http://www.apache.org/mirrors/ Note that many of the mirrors of OOo and of apache also mirror large amounts of other open source software. OOo uses mirrorbrain [1]. Apache uses some CGI scripts to accomplish a lot but not all of the same functionality. The set of data ASF pushes to its mirrors is somewhere downwards of 50GB [4]. The set of data OOo pushes is somewhere downwards of 300 GB [2]. It doesn't seem like a good idea to push an additional 300GB to all existing apache mirrors: that isn't quite what they signed up for. As I understand it Oracle wants to transition out of maintaining the OOo infrastructure eventually. So if OOo gets accepted into the incubator, it seems like the smart approach would be to duplicate the existing OOo mirrorbrain installation onto some apache hardware, with the people that look after the OOo and apache mirrors figuring out the specific details. Will be a fair bit of work I imagine so definitely needs volunteers to step up and all that, but nothing particularly scary I think, assuming the existing OOo mirror maintainers [3] help out. Without their help it will be much harder to figure out how to do things! If most of the people that worked on mirroring at OOo are now at TDF (and looking at mail archives it seems that might be true [7]), better be extra nice to them TDF folks :-) Long term, if there's people to do the work, one could imagine updating the custom ASF mirror infrastructure to use mirrorbrain which seems like a cool tool...but that is a _lot_ of work because the existing CGI scripts integrate into the download pages of every apache project. cheers, Leo PS: LibreOffice also uses mirrorbrain [5], having about 65 mirrors. They required only 15GB for a mirror though [6], not the OOo footprint 200GB. Sounds much more reasonable... [1] http://mirrorbrain.org/ [2] http://wiki.services.openoffice.org/wiki/New_OOo_Mirror_Structure http://distribution.openoffice.org/files.html http://wiki.services.openoffice.org/wiki/Mirrors_Project [3] http://openoffice.org/projects/distribution/lists [4] http://www.apache.org/info/how-to-mirror.html [5] http://download.documentfoundation.org/mirrors/all.html http://wiki.documentfoundation.org/Mirrors [6] http://download.documentfoundation.org/mirroring.html [7] http://blog.gmane.org/gmane.comp.documentfoundation.mirrors --------------------------------------------------------------------- To unsubscribe, e-mail: general-unsubscr...@incubator.apache.org For additional commands, e-mail: general-h...@incubator.apache.org