On Tue, Aug 25, 2009 at 10:55 PM, <johansen at sun.com> wrote: > On Tue, Aug 25, 2009 at 10:13:05PM +0100, Peter Tribble wrote: >> I'm using the hardware available to me. Generally that means >> more modern hardware for OpenSolaris. It's not an absolute >> number, but a straight comparison on identical VirtualBox installs >> on my home Ultra 20 gives: > > More specification, please. ?The Ultra 20 is an EOL'd system.
Which is completely irrelevant. I'm not comparing it to something else (although if you want me to do that, I'll note that my Ultra 60 running SXCE was faster at packaging that the modern core 2 laptop running OpenSolaris that I used recently). The point is that with identical configurations on the same hardware, pkg(5) is much slower than SVR4. >?It's > possible to make benchmarks look arbitrarily poor by manipulating > operating system tunables and fiddling with the hardware config. Come now. I've installed the system and left it untuned. >> That's for S10; SXCE is generally quicker, and can easily be made >> significantly faster. The SVR4 numbers are pretty poor, really. Essentially >> every time I've compared performance, pkg(5) comes out several times >> worse. > > If you're comparing S10 and SXCE to OpenSolaris, you're comparing apples > and oranges. ?This isn't really a fair comparison. But that *is* the comparison of interest. >> My install timings on an old sparc box come out at about 44 minutes >> for SXCE, and 70 minutes for half the software for OpenSolaris. > > Pkg(5) is still under development. ?It's rather unreasonable to suggest > that the unoptimized performance today, in development builds no less, > will reflect the final performance of the product at release. ?We have > more features to add and more optimization to perform. Pkg(5) has had considerable development effort expended on it. It's been released in shipping product for over a year. SVR4 has been neglected and left to rot. > If you'd like to help analyze and fix, we're happy for your > contributions. ?I'm willing to consider reasonably crafted patches for > well identified performance problems. ?However, time I spend answering > e-mail is time that I'm not spending writing code. Are you willing to accept that pkg(5) does have serious problems with performance, and needs improvements of an order of magnitude or more? -- -Peter Tribble http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
