Solaris 11 Released 224
angry tapir writes "Oracle has updated its Unix-based operating system Solaris, adding some features that would make the OS more suitable for running cloud deployments, as well as integrating it more tightly with other Oracle products. While not as widely known for its cloud software, Oracle has been marketing Solaris as a cloud-friendly OS. In Oracle's architecture, users can set up different partitions, called Zones, inside a Solaris implementation, which would allow different workloads to run simultaneously, each within their own environment, on a single machine."
Re:8 char usernames (Score:2, Informative)
You can have longer than 8 character user names, but the characters after 8 are ignored. It's defined in limits.h as LOGNAME_MAX. It's an ABI restriction, hard-coded in several binary formats, NIS restriction, and UNIX interoperability issue. Another limit is the 32-bit character limit from POSIX, but that's been removed, I understand. Don't blame me--I'm just telling you.
Re:Cloud hosting (Score:0, Informative)
Actually, Linux tends to be faster than Solaris even on SPARC hardware, let alone x86.
http://www.stdlib.net/~colmmacc/2006/04/13/more-ubuntu-on-t2000/
Re:I guess Ellison changed his mind (Score:5, Informative)
Quite the opposite. In your own link he summarized by saying:
"I'm not going to fight this thing." but "I don't understand what we would do differently in the light of cloud computing, other than change the wording on some of our ads."
And sure enough, their ads now show how great Solaris is for cloud computing. Based on what?... zones, which have been in Solaris for a number of years.
Re:ZFS v31+ at last? (Score:4, Informative)
While ZFS at v28 has proven to be both a lot of fun and very useful for many of us, the updates since (first available for general use with Solaris 11 Express last year I believe) add a few really nice features
Careful, they've also abruptly removed a few really nice features in later versions that have caused major headaches for me and many others. For example the "aclmode" property was completely removed from version 31 - completely breaking a lot of deployments that made extensive use of ACLs. Version 33 released today with Solaris 11 thankfully restores that feature after significant outcry from affected customers (I believe Illumos went forward and restored it on their own as well) - but the damage has been done in a lot of cases.
Just a word of warning to be very careful before running "zpool upgrade" as Oracle's philosophy on backward compatibility and stability of existing features seems to be quite different than that of Sun.
Re:$1,000/year per CPU for non-Oracle hardware (Score:3, Informative)
Ever since Oracle bought out Sun, they went overboard with the licensing costs for Solaris. Remember a few years back when Sun will let you run Solaris 10 for free? Well no more, if you have a non-Oracle two processor server it will cost you $2,000 per year. You don't own a license, you are basically renting the privilege to run Solaris on a server for one year. Also, you only get one flavor of support which they laughably call "premium". Their support is a joke now, and in my experience the good Sun engineers left a long time ago. For starters, you now get to talk to an overseas helpdesk which logs your call and for severity one issues, they give you a call back in an hour (if you're lucky). It used to be you will call an easy to remember number (1-800-USA-4SUN) and you will get a live transfer to a knowledgeable engineer to fix your problem. A few years ago I used to be a staunch supporter of Sun and Solaris but it seems like Oracle has done everything to drive me away from Sun's hardware and software. I am pretty sure I am not the only one either.
I don't know where people are getting this $1000/socket bullsh*t. Maybe that's some ridiculous list price, but unless you're a moron, you won't pay anywhere close to that for full HW and OS support on Sun/Oracle hardware. The last time we renewed our support, I believe it was in the realm of $400-800/yr for HW/OS support on our x86 servers [dual socket Opterons and quad-socket Xeons]. The SPARC servers were a bit more expensive, closer to $2000 for support on a T5240 [dual-socket 8-core x 8-thread/core T3+ CPUs]. Remember, that includes HW support, fans, HDs, RAM, CPUs, motherboard replacements, whatever with same-day onsite service [well, in theory, in practice it's often the next day, but most of our hw failures aren't critical to our services so we don't push them very hard].
That's not to say I love Oracle's support since the buyout. Though HW failures are typically handled fairly quickly, their support website is a nightmare, and getting an IDR [Interrum patch] on anything less than a major OS bug can be a long-term process, but I'm not sure it's significantly worse than any other vendor's support in the long-run.
Re:Nothing to see here (Score:4, Informative)
It's Linux. If you need it, build it.
Arrrgghhhhh! Grow up!!!! That is like dumping someone on a plot of land and telling them if they need a house they should just build it themselves. Not everyone is an architect, master builder, plumber, electrician etc. etc. Nor is everyone capable of writing their own file system software. This argument is not sane and I cannot believe that supposedly intelligent people have continued to make it for many decades now.
Or are you just bored and trolling?
Re:*crickets chirping* (Score:4, Informative)
FreeBSD's ZFS is years behind what Illumos offers in features, and shows no signs of catching up.
FreeBSD 8-STABLE and 9-RELEASE contain ZFS v28, the same version of ZFS as OpenSolaris. iXSystems is now funding development, and it has seen quite a lot of bug fixes that have yet to be back-ported to any Solaris version.
the FreeBSD implementation is still dogged by performance issues. Any significant workload on ZFS is still marginal compared to, well, pretty much anything else (including, dare I say, NTFS on Windows).
I installed FreeBSD 9 BETA on a machine with three disks in a RAID-Z configuration and the only time the bottleneck for reading and writing to the array was not the GigE connection, was when I was writing to a compressed deduplicated filesystem. Then the CPU was the limit, at about 20-30MB/s. That's with a pretty anaemic CPU (1.6GHz AMD Fusion) and with WITNESS turned on in the kernel, which adds lots of extra error checking around kernel code and slows everything down.
Re:Zones (Score:5, Informative)
Given that Zones can have:
different login identities
different network interfaces
different hostnames
different hardware available to them (disks, adapters, etc.)
be configured to use resource pools thus different amounts of cpu, floating or fixed
Yes, I'd say they are much more useful than chroot.
Re:Cloud hosting (Score:5, Informative)
There is no modern OS that can mitigate an application that is bad
Of course there is! That's the entire point of the OS. If an application can bring down the OS, then that's an OS bug. The responsibility of a time sharing system is to ensure that no process and no user monopolises the resources to the extent that others are unable to do anything. The correct behaviour in this case (and the behaviour I've seen on Solaris, recent OS X, and FreeBSD), is for the Apache process to slow right down and other users to experience a noticeable amount of degraded performance (unless they're running with elevated privileges). Being unable to log in from the console because of the actions of an unrelated userspace process is simply unacceptable.