Solaris 11 Struggles Against Linux Distributions

[kgardas]

Comparing 32bit Solaris app code to 64bit Linuxs is not fair!

Hi,
just to defend Solaris a little bit. This OS's gcc compiler by *default* produces 32bit applications even on amd64 hosts and even Solaris 11 is 64bit. On the other hand majority of Linux distros for amd64 are pure 64bit which means also their gcc compilers produce 64bit apps. As amd64 is way much better in perfromance than pure old x86 I would really strongly recommend to rerun the benchmarks and enforce 64bit app compilation on solaris by using `-m64' switch. Nothing more is needed.
Thanks!
Karel

[uid313]

I still wonder from time to time what would've happened if the OpenSolaris code had been released as GPLv3 as some rumors had speculated.
Solaris had some interesting tech in it, but overall it was very hardware-limited. You had to basically buy hardware for it. Linux was already miles ahead of it in that regard.

Any chance Solaris had to shine is long gone. Oracle was just the final nail in the coffin.

Yeah, and wonder what would happen if it was under released under the Linux-compatible GPLv2 license, or the 2-clause BSD license or ISC license.

[Mystro256]

suspicious indeed

The SciMark Monte Carlo results look suspicious. All that should be doing is drawing random numbers and it wouldn't be using OS-features like cryptographic random devices to do so either.

Why is Linux (CentOS) performing substantially worse than Linux (Ubuntu)?

The Solaris results are also unusual in that it performs poorly on tests that have nothing to do with the OS. I'd be really curious about performance if benchmarks were compiled by suncc. However, that wouldn't be a fair comparison.

I agree, that did look odd. I also would like to add, why is Fedora noticeably lower than Ubuntu? The gap seems suspicious or perhaps biased, unless of course it's a regression/bug caused by GCC 4.7.

[vik1]

Flawed Test Results

Hi, these tests results are flawed. Solaris is using ZFS in these tests, so to accurately compare it with other Linux distributions you would have to partition the Solaris filesystems with UFS or change the Linux distributions to use BTRFS. ZFS is known to be slower than UFS due to software RAID checksums but gives you a lot more flexibility (snapshots and repartitioning) and redundancy.

[kallisti5]

I was getting behind the whole OpenSolaris thing... there are a *few* cool features such as crossbow Sun was starting to develop in the open.

However Oracle and their general taint made me drop all interest pretty quickly.

[kraftman]

Yeah, and wonder what would happen if it was under released under the Linux-compatible GPLv2 license, or the 2-clause BSD license or ISC license.

Nothing would happen. ZFS would be just ported to Linux. That's all.

[kraftman]

Hi, these tests results are flawed. Solaris is using ZFS in these tests, so to accurately compare it with other Linux distributions you would have to partition the Solaris filesystems with UFS or change the Linux distributions to use BTRFS. ZFS is known to be slower than UFS due to software RAID checksums but gives you a lot more flexibility (snapshots and repartitioning) and redundancy.

Tests aren't flawed in this case, because to have some equivalent to UFS Linux should be using Ext2... If Solaris does only have two file systems it's not Linux problem. btrfs isn't stable yet, so your suggestion is stupid.

[kraftman]

Hi,
just to defend Solaris a little bit. This OS's gcc compiler by *default* produces 32bit applications even on amd64 hosts and even Solaris 11 is 64bit. On the other hand majority of Linux distros for amd64 are pure 64bit which means also their gcc compilers produce 64bit apps. As amd64 is way much better in perfromance than pure old x86 I would really strongly recommend to rerun the benchmarks and enforce 64bit app compilation on solaris by using `-m64' switch. Nothing more is needed.
Thanks!
Karel

To defend or to burden it? 32bit should have less overhead.

[vik1]

Tests aren't flawed in this case, because to have some equivalent to UFS Linux should be using Ext2... If Solaris does only have two file systems it's not Linux problem. btrfs isn't stable yet, so your suggestion is stupid.

Wow, troll much? I suggested 2 different, but more accurate, ways to test Solaris vs. Linux and your comment is "that's stupid." Take a look at what software raid does to performance (btrfs vs. ext4 for instance) and then try again, genius.

[johnc]

No, it was Linux. This is the reason why SUN is dead.

Well, Intel played the largest role.

It's not like customers were buying Solaris.