> Dear all: > > Will Solaris 10 perform better than Solairs > Express? > For instance: 1) Solaris 10 update 3 vs. Solaris > Express B50, which one performs better?
Sorry for telling the truth: Solaris 10/SunOS5.10's kernel is less cpu-hungry, than Solaris 11 aka SXCR/SXDR/Nevada/SunOS5.11's is. Both on SPARC and on x86/x64. Every sane person will notice that, I do not trust those benchmark figures once posted to performance-discuss. Just test Sol10 versus 11 under QEMU or on any sub-750MHz UltraSPARC machine. I'm not talking about all the fat "thin" clients in userland, such as Firefox. Just about basic "feeled" performance (even in text mode, doing tryouts in /tmp to reduce/exclude I/O factors). I stuffed 2GB mem into a SunBlade-100 and wanted to replace my 24x7x365 Blade-2000 with it in order to save electricity and save on huge power costs. But That Blade-100 (which once feeled really quick under Sol8 and 9) is just unusably slow under SXCR. Solaris 10 was the last one to be usable there. And I'm by no means a "power gamer user". My needs/standards are very moderate (as this is not a compile machine, but just serves as Desktop/mp3/mail/web-browsing machine). But Nevada's performance does not even meet them. I'm not skilled enough in the field of kernel performance analysis (still didn't read the second book of Solaris Internals). Could somebody from the kernel group give me an answer please? What is to be blamed, the newer Studio compilers (10 and now 11), or has complexity of kernel's tasks/idle workload really grown that much? A gcc built kernel feeled a bit quicker, but I have no proof, no numbers on that. I already noticed in the nineties under MS-winDOS-9x. If an upgrade to an application came built by never compilers it would have taken up 20 minutes on my 486DX 33MHz just to load that program. Otherwise looking like built from same source base, as previous version, respectively. Is this the price for having new ISA-upgrades in the compilers? Why? No alternative method? Sorry for unsophisticatedly talking about "feelings" here. But you could verify those "findings" by just measuring it on a stopwatch. Also I noticed that kernel/system- cpu time get high very fast. Thank you. I'm not complaining. Just reporting an issue in the hope you might succeed in improving things a bit. p.s. I wrote this via jive, as I'm still not re-subscribed to the generic list. I chose "CC Forums:" performance discuss. This might not be refelcted in your CC: field, pls. add that. %martin This message posted from opensolaris.org _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
