On Fri, 2004-04-30 at 04:33, Farkas Levente wrote: > hi, > is there any chance to improve yum performace? I don't know the current > algorithm, but the current dependencies resolving is very slow? and as > the number of repositories and package number are growing it's geting > slower and slower. is like in the old XT world (when we has to do some > progress bar like the .... in yum). wouldn't it be possible to > store/generate some kind of additional information during the repository > generation (when the time is not so important) in order to speed up the > online usage of yum? or rewrite the most time consuming part to native. > or ...? > yum is a realy nice tool, that's what we waiting for since rpm, but it's > slow:-( > yours. work is being done to improve performance, yes. I'm hampered by other work I have to do so it is slow going. Right now there are a number of things left to do but yum-HEAD is sliding, slowly, toward more functionality. Once I get it back to where it is in 2.0.X for functionality I'll make an alpha release of that tree and I'll be focusing on optimizing all the new stuff. I want yum to be fast for day-to-day use and I think the places that suck up the most time now are being addressed. downloading all the headers to install a single rpm is no longer necessary - it just downloads a header of a package that is actually being installed. Dep resolution appears to be a lot faster now and after I'm happy with how it is working I'm planning on writing some routines to pre-resolve package deps before going into rpm at all. Memory use should be a lot lower now, too. The trick will always be this, though: The more repositories you use, the more memory and time yum will require. It has to traverse all of those packages in all of those repositories and that takes time and memory. So if you have 10 repositories in your yum.conf with 2000 packages each, expect to wait. -sv
