OSNews

From what I hear hardware support is spotty. Similar to Linux in '95. So check to make sure your stuff is supported before you start. Otherwise, I agree with you, Solaris sure looks amazing.

Hardware support is vastly better than where Linux was in '95, though still not up to what Linux is currently.

hmm i guess if you only use Solaris as a toy, its useless without all the bells and wistles, but people that use there system as a tool, will still find lots of uses for ZFS, you can still store all your data on ZFS users home directories, databases, web documents just fine even though it doesn't support root on ZFS.

ZFS makes its incredibily simple to upgrade your system and keep the data that is on a ZFS pool. Just install/update as usual, and then

zpool import

it will list the pool availible, then zpool import poolname ; done

Good things come to those who wait, I'm sure ZFS Mountroot is coming soon. Maybe you should read the document I linked to to see what you can do with ZFS despite not having ZFS root.

Yes Gnome 2.6, and Mozilla are outdated, and suck. If you want Gnome 2.14.x and firefox, then use the latest Nevada release of Solaris. It might also be possible to update Gnome and Mozilla/Firefox by using JDS vermillion from the Open Solaris site.

Hm, that's alpha software. I'm not planning on using this even if it's from SUN.

Yes Gnome 2.6, and Mozilla are outdated, and suck.

I think that's rather subjective to say. Some users may dislike those versions, certainly, but it's not like users of Debian stable can say much worse about their versions of software

Seriously, if it does the job, many people are perfectly content. Just like people still running Windows 3.1. If it does what they need, who are you to say "it sucks"? The more accurate thing to say would be, "I dislike using GNOME 2.6 and the version of Mozilla shipped with Solaris 10. I wish newer versions were included."

No it is not subjective. Using Solaris as my desktop for many years, Gnome 2.6 looked and worked a treat when it first came out. Now for me, the old libraries were a daily problem, and it wasn't "doing the job". Thanks to the work of Sun's desktop team, people can now use the much more stable 2.14 verion of Gnome. Having used both I can truly say that 2.6 sucks big time compared to the latest version. If you or any other one else only require Gnome 2.6 or Windows 3.1, then all the best.

No it is not subjective.

Yes, your opinion, and mine, are subjective things. As much as we would all like to believe that are opinions are facts, they are not.

The SDLC issue has been resolved-- we had a software configuration error, which has now been fixed. Sorry about that!

Why would you want to run ZFS in userspace? The latency must be terrible.

Because that's the only way Linux will see ZFS, short of a license change by Sun, and having some compatability is better than having none.

Edited 2006-06-26 20:03

"Because that's the only way Linux will see ZFS, short of a license change by Sun"

Or Linux could change its licensing.

"Or Linux could change its licensing."

Unlikely. Unlike Linux, Solaris copyrights, all or most of it is retained by a single entity and there has been some talks about moving into GPLv3

http://www.computerwire.com/industries/research/?pid=99BC40E4-089B-...

My point was that the original poster made it sound as if Sun were exclusively at fault. They're not. Either side could change their licensing to be compatible with the other.

Sun is exclusively at fault. Linux has been out there and GPL for a decade. Sun had every chance to choose a compatible open source license had they wanted to. Unfortunately, Sun didn't want to be compatible and wrote their own. So yes, it is exclusively Sun at fault here.

This, by the way, is not saying anything against Sun's license, it's simply pointing out that Sun deliberately chose to be incompatible.

Sun is exclusively at fault. Linux has been out there and GPL for a decade. Sun had every chance to choose a compatible open source license had they wanted to. Unfortunately, Sun didn't want to be compatible and wrote their own. So yes, it is exclusively Sun at fault here.

Wait a second.... Isn't this a piece of Sun's technology we are talking about here. If Linux wants features it has to be more open about it's licensing. CDDL has no problems interacting with any other license except GPL.

This, by the way, is not saying anything against Sun's license, it's simply pointing out that Sun deliberately chose to be incompatible.

Sun's CDDL is almost compatible with most licenses it is GPL that is not compatible with other licenses. Think about the manyn projects like Mozill for example that had to dual license to keep the GPL zealots happy.

I certainly didn't mean to imply that anyone was at fault.
I was merely pointing out the way things stand.

It's not Sun's fault that the GPL didn't suit their needs for OpenSolaris, nor is it Sun's fault that the GPL prevents the CDDL'd code form being used in Linux.

lol

Dear all-linux-contributors-since-1991,

Please confirm your assignment to me of copyright for any and all code modeifications you made to the 'linux' project in writing asap. This is neccesary so that I can relicense it under a license you never subscribed to, in order to aquire a FS implementation from Sun. As you are probably aware without this important action Linux will surely be losing market share to this widely adopted free OS from these Sun guys.

Regards, Linus.

haha, I see the irony here. I just hope this isn't as big an issue when Linux goes to GPL v3.

I love using Linux and I love using Solaris. I also like the GPL v2 and I like CDDL. They both have their uses and their place. Sun faced a lot of issues with open sourcing Solaris. CDDL was implemented to get around those issues. This has been discussed to death when Open Solaris was initially released. I have no intention of chewing the proverbial "cud".

I hope there comes a happy meeting point between these operating systems to share code and features. The end users are the real beneficiaries if this happens.

I like using ZFS and other Solaris features. A lot of them are very easy to use and implement. Zones, smf manifests come to mind along with ZFS. There is a learning curve attached to DTrace but it is well worth the effort as the gains are immense.

I like different operating systems mainly because I like to learn. I like to play around with different features and check out the results for myself. That was the spirit behing the Linux wave IMHO. Curiosity and desire to learn about how operating systems work. I hope the FOSS community never forgets this reason.

EDIT: Grammar mostly

Edited 2006-06-27 09:52

I don't think that the Linux kernel will use the GPLv3
http://trends.newsforge.com/article.pl?sid=06/01/25/238257&from=rss

>Why would you want to run ZFS in userspace? The latency
>must be terrible.

As pointed out, compatibility is better than none. Licensing issues stand in the way of in-kernel ZFS for Linux at this time.

As for FUSE performance, it's better than you might expect.

ZFS is an interesting and potentially useful file system, but it's not all things to all people. I suspect a limited subset of users will have a legitimate use for it. To others, it will be a break from experience will offer no compelling features that are worth migration costs (time, mostly).

That said and all... thankfully Sun has finally delivered it. Good show, and about damn time after shouting at the heavens about it for nearly two years.

Edited 2006-06-26 20:12

I think it has applications in many areas. It's a lot easier to manage than LVM and offers a lot more features (compressions, end-to-end data integrity, self-healing data). I know a lot of HPC centers are interested in it for that reason.

Also, it's rather ironic that you'd dismiss ZFS as too different from existing filesystems. There's such regular upheaval of Linux subsystems that I thought most Linux admins would be used to change by now.

>It's a lot easier to manage than LVM and offers a lot
>more features (compressions, end-to-end data integrity,
>self-healing data). I know a lot of HPC centers are
>interested in it for that reason.

Someone who has been using LVM for years isn't going to find it a lot easier to manage, was my point. Many features of ZFS are great - and there may be no parallel to them for some time (though some features, there are existing matches), and I am not denying that - at least, not now that Sun has actually delivered the product.

But are they worth the migration? Everyone who is interested in ZFS' features will have to answer that for themselves. It's not going to change the lives of those without large arrays of disks. Indications are it's a good FS with good tools and storage abstraction, and that will only excite a select few end users.

>Also, it's rather ironic that you'd dismiss ZFS as too
>different from existing filesystems.

If you could point out where I dismissed ZFS, be my guest. It seems you are seeing what you want to see.

>There's such regular upheaval of Linux subsystems that
>I thought most Linux admins would be used to change by
>now.

Perhaps... or perhaps not. I've standardized on two file systems on Linux systems and have used them to good effect. SANs have their own software which is somewhat similar to ZFS' disk management.

It's not going to change the lives of those without large arrays of disks.

Spoken like a person who doesn't understand ZFS' snapshot and clone features. Think backups and home user. Two disks create a mirror. Snaphot important work with one command instantly.

Edited 2006-06-27 00:12

>The reason you won't get the point is you are someone
>who has been trolling on Sun related post for years.

It would be convenient to label me that, wouldn't it? But, it's not the case. I just disagree with some of your ideas, and you don't care for that.

> Spoken like a person who doesn't understand ZFS'
> snapshot and clone features.

But I do. Useful features, but using the file system directly isn't the only way to achieve similar functionality. I've yet to see how those operations affect performance, but we'll leave that to another discussion, as I've not been able to test that in the real world since ZFS was just released.

For you, this may be the ultimate set of features. For me, well... I like to store my backup data off-site. People have different needs... see?

> Two disks create a mirror.

To quote YTMND... O RLY?

> Snaphot important work with one command instantly.

Handy, but not revolutionary. Nor unique. VMS and NTFS have similar functionality. Many other file systems could have this added into them as a future feature, it's essentially a spin on revisioning.

> As their disks get bigger and bigger, they can easily
> add more storage to the pool and just grow it
> endlessly and "never" run out of disk space, never
> have to move a file again, never have to copy
> partitions. Just keep slapping drives willy nilly
> onto the system, and it gets bigger -- just like RAM.

ZFS is not the only answer to that, though. But credit where it is due, it's a very good and potentially the simplest implementation of the ideas.

>But it sure makes anyone who happens to be running Sol
>10 already a good reason to upgrade, even without ZFS
>root.

I see it as more of a forklift upgrade situation. When it's time for the forklift, it may be time for ZFS. But for new installs which will be accessing plenty of disks, by all means use ZFS if Sol10 is your platform of choice (maybe even if it isn't, depending).

But I do. Useful features, but using the file system directly isn't the only way to achieve similar functionality. I've yet to see how those operations affect performance, but we'll leave that to another discussion, as I've not been able to test that in the real world since ZFS was just released.

ZFS has been a part of Solaris Express for a while... you could have installed it and used it if you had so wished.

For you, this may be the ultimate set of features. For me, well... I like to store my backup data off-site. People have different needs... see?

I am tired of your constant trolling on Sun related stories so I am going to stop feeding you. Ciao.

Spoken like a person who doesn't understand ZFS' snapshot and clone features.

For those who are used to LVM, RAID and SAN hardware and software out there it isn't exactly going to rock their world with something new. Integration into the filesystem might be slightly more convenient, but new features? Hardly. You hardly need it all integrated directly into the filesystem.

Think backups and home user.

I hardly see the home user as being a user of ZFS and Solaris.

Two disks create a mirror.

Wow. RAID anyone?

Snaphot important work with one command instantly.

I do that now. VMS did this, and anyone with LVM will be doing it in order to make these things called *off site backups*.

I hardly see the home user as being a user of ZFS and Solaris.
I have almost totally converted (bar a small ufs grub partition) to ZFS on my single disk laptop. With the features such as snapshots, compression, I (as apple has) can see are a huge bonus for a home user.

For those who are used to LVM, RAID and SAN hardware and software out there it isn't exactly going to rock their world with something new. Integration into the filesystem might be slightly more convenient, but new features? Hardly. You hardly need it all integrated directly into the filesystem.

From this I can assume that you have never actually used ZFS.

For those who are used to LVM, RAID and SAN hardware and software out there it isn't exactly going to rock their world with something new. Integration into the filesystem might be slightly more convenient, but new features? Hardly. You hardly need it all integrated directly into the filesystem.

you are mistaken here. Veritas the #1 seller of LVM and highend filesystems are very scared have you noticed that they now offer VxVM and VxFS both free on low end systems is this is a coincidence? I don't think so, licenses for both of these were $3,000 each.

The basic parts of ZFS that do Raid 1( mirror) and Raid 0 (concat) are interesting because they do checksumming of all data, from the OS to the hardisk and back. No other filesystem does this, and this ends silence data corrruption. Its bad enough when a drive or controller dies and takes your data. What really hurts is when a drive or controller is silently dieing corrupting your data as it degrades. Of course you are being dilligent and backing up your data. With any other filesystem. You would be backing up corrupt data and this could cost you data, that is possibilty irreplaceble because you backed up corrupt data.

Now lets look at what really has hardware storage companies rethinking there product line. ZFS does both raid 5 and raid 6. Well not quite they do something like it. RaidZ and RaidZ2. It has all the same features as raid 5 and 6. But writes are not Slow. Because ZFS uses a variable block size, all writes are full block writes. So no expensive battery backed up ram or nvram is needed to keep the write cache if there is a power failure.

if this wasn't enough. There is more. ZFS has snapshots. Yes snapshots are nothing new, but the way ZFS does them is economy changing. Up till now snapshots required commercial software packages read $$,$$$ for the cheapest, or required you to take your storage off line. Or special hardware to accomplish, or both hardware and software. ZFS does it instaneously and for free as in no downtime, or extra costs.

For a complete comparison of ZFS vs. Linux lvm and raid features check out http://unixconsult.org/zfs_vs_lvm.html

If it was just these features it would change the enterprise storage world. ZFS does more. It creates a user interface that is simple enough that you can walk your grandmother through setting it up over the phone.
so no more studing 500 page manuals, and a phd in storage-ology needed to make it do your bidding as shown at http://uadmin.blogspot.com/2006/05/why-zfs-for-home.html

I could continue with even more features that change the world. But i will save those for later posts.

Solaris still needs to catch up. Partially it is excellent - but partially it has very old cruft. For example their /bin/ksh is still ksh88 while Linux, FreeBSD, OSX/Darwin all use ksh93.

There is already a project in progress to integrate ksh93. As for the rest of the software, there's a reason it isn't "bleeding edge." SUN has a very long support cycle, and their customers want certain guarantees that updating or upgrading their system won't break anything. They tend to take a view of "we ship it, we own it" when it comes to the core OS distribution. Things that are on the companion CD can be updated more frequently, but still have certain restrictions.

So, it isn't accurate to say it has a lot of "old cruft." Remember that Solaris is one of the last true UNIX operating systems left, and as such, there are many requirements that exist for it be certified under SUS (Single Unix Specification) that don't exist for other operating systems (such as Linux).

While this may be detrmiental to those who enjoy bleeding edge software, those of us that prefer well documented and stable systems enjoy it as it is.

> There is already a project in progress to integrate
> ksh93.

Are you referring to http://www.opensolaris.org/os/project/ksh93-integration/?
Sun canceled half of that project. The only part "left" is to add ksh93 to a weired location where no one will find it.

Today is the year 2006. SUN HAD 13 YEARS to update their /bin/ksh shell. And did NOTHING. And even is refusing cooperation with the open source community to get /bin/ksh updated.
Today portability is a major issue and /bin/ksh makes portable shell scripts impossible. And backwards compatibility to a /bin/ksh which dumps core, has many bugs and was even abandoned by it's authors is ridiculous.

Yes, Solaris is one of the last Unix versions. Maybe it's time for customers to abandon the ship and switch over to a real open source operating system where a software update cycle doesn't take 13 years.

I just say: Thirteen YEARS.

This cool example how Sun treats customers was brought to you by Sun service

Maybe you should reference "The New Kornshell Command and Programming Language" by Morris I. Bolsky and David G. Korn, starting on page 328 is the quick reference showing features for the korn shell. The areas that are highlighted in gray are those features that are only supported by versions later than 11/16/88. While there are some features that would be nice (such as printf), I don't see them as show stoppers. Compared to the other guys I work with who use tcsh, I am "old skool" and use ksh.

I think you are making a mountain out of a molehill.

Today is the year 2006. SUN HAD 13 YEARS to update their /bin/ksh shell. And did NOTHING.

May be this point is lost on you. It is not Sun but Sun's customers that dictate these things. If Sun changed every thing willy nilly like linux distros used to do, customers would be pissed.

Say some incompatibility in ksh93 breaks existing scripts because the customer updated thier OS, they call Sun support. Which is expensive for Sun. Sun and most vendors to enterprises usually announce major changes one or two release cycles in advance and put the new component in an alternate place so cusotmers can migrate thier old stuff off production.

I can garuntee Redhat won't be updating RHEL AS ever time a minor version of some tool is released. That is why fedora exists.

Yes, Solaris is one of the last Unix versions. Maybe it's time for customers to abandon the ship and switch over to a real open source operating system where a software update cycle doesn't take 13 years.

I can bet that most Enterprise customers who have deployed linux aren't running the latest linux 2.6.xx kernel. Most are probably in 2.4 or even 2.2.

I seriously doubt you have had any exposure to enterprises or data centers.

BTW there are still shops running extremely old Sun boxes in production, some even older than 13 years. Data centers and Enterprises don't like change too often. They have billions riding on each change and can't afford for things to go wrong.

Compatibility is a good thing, but I think in the case of /bin/ksh Sun has chosen the dead end and seems to stick with this option.
I've been working in a datacenter for almost twenty years now and have to admit that while Solaris is nice it lacks something called "INTEROPERABILITY". It is not possible to write a shell script on Linux or OSX and use it on Solaris when /bin/ksh is involved because there is a difference of almost 15 years in development between both ksh88 and today's ksh93q+.

Interoperability has become a key factor in datacenter deployments and Sun is unable to deliver that, comparing to Linux which is almost the number one in interoperability thanks to the fact that they implement almost all protocol stacks, shells, standard APIs and other goodies.

Sticking the "Open" in front of "Solaris" did not change that either. So far Sun still owns us (the customers and the stockholders) the proof that their strategy has changed and now explicitly includes "interoperability". Not only the kind "interoperability" of the past which included that each single feature needed to be paid for by customers - I mean "interoperability" which Sun implements BEFORE customers hit them. Features like ZFS are cool but this is still proprietary and do not include options for interoperability.

Sun still has to learn what "Open" really means. And I hope that Mr. Schwarz and his crew learn this lesson soon, otherwise the next round of layoffs will me harder.
And the apparent failure of the ksh93 integration project will be one of the topics of the next stockholders' meeting. Something is seriously going wrong at Sun and this needs to be fixed.

Today is the year 2006. SUN HAD 13 YEARS to update their /bin/ksh shell. And did NOTHING. And even is refusing cooperation with the open source community to get /bin/ksh updated.
Today portability is a major issue and /bin/ksh makes portable shell scripts impossible. And backwards compatibility to a /bin/ksh which dumps core, has many bugs and was even abandoned by it's authors is ridiculous.

Pardon, but this is hardly true. I really don't like it when people post lies. SUN has not been "refusing" to update to their /bin/ksh shell. It was considered, and it was talked about. However, as a business they have the ultimate right to determine what happens based on what they believe to be their customer's needs.

In this particular case, they believe that backwards compatability is a core customer need. As such, they are not replacing the base ksh (at this time, so far) with an incompatible version of ksh.

In addition, since the project has not yet been integrated, it is unreasonable to claim what will happen before everything has been decided.

Since you are not one of the contributors to the ksh integration project, I find it difficult to value your complaints.

Pardon, but this is hardly true. I really don't like it when people post lies. SUN has not been "refusing" to update to their /bin/ksh shell. It was considered, and it was talked about. However, as a business they have the ultimate right to determine what happens based on what they believe to be their customer's needs.

It was talked about: Yes
It was investigated: No

I don't even dare to ask whether it was really considered. The issue was simply put aside with the comment that both shells are not "compatible" (and this accompanying commentary is from Henk Langeveld, the Sun engineer who was lobbying for the /bin/ksh upgrade for many years).

There was no project until 2006 which actually looked at the details and differences between ksh88 and ksh93, twelve years after the first request to update /bin/ksh.
And again Sun has chosen the "minimum" path: They accept ksh93 in Solaris, but only as /bin/ksh93 and not as /bin/ksh. This is why I call this project a failure.
It fails I archive an improvement in interoperability which Solaris needs more than everything else.
The days where customers have Solaris/SPARC-ONLY installations are over and customers need a minimum level of interoperability, something which Sun is currently unable to deliver.

It was talked about: Yes
It was investigated: No


You have no proof of that, sorry.

I don't even dare to ask whether it was really considered. The issue was simply put aside with the comment that both shells are not "compatible" (and this accompanying commentary is from Henk Langeveld, the Sun engineer who was lobbying for the /bin/ksh upgrade for many years).

There was nothing simple about it, and I know for certain that it was more than just "put aside."

And again Sun has chosen the "minimum" path: They accept ksh93 in Solaris, but only as /bin/ksh93 and not as /bin/ksh. This is why I call this project a failure.

You seem to miss that SUN's primary concern is that scripts written for Solaris that depend upon the current /bin/ksh MUST continue to work without failure. ksh93 IS NOT 100% backwards compatible with ksh88. Therefore, they chose not to put it there. You also failed to mention that there have been discussions about what it would take to make a version of ksh93 shipped with Solaris fully backwards compatible with the current version that is shipped.

Some people value new features over compatability. SUN values their customers being able to depend upon the software they ship. They have no desire to break customer scripts written for the current /bin/ksh. That is commendable and is something highly lacking in the Linux world.

You can view the project however you like, the fact that it will be integrated is a great thing and a step in the right direction. Whether you view it as a failure is up to you, I consider it a success.

It fails I archive an improvement in interoperability which Solaris needs more than everything else.

That's rather subjective. Interoperability beyond what it already has today is not something I'm concerned about. Quite frankly, it's almost laughable to be talking about Linux and interoperability. Linux doesn't follow the UNIX standard, Linux is NOT Unix. The Linux world enjoys breaking people's software with the excuse that source compatability is good enough, and if your software was open source it wouldn't be such a big deal.

Customers like me want a reliable, efficient, and powerful system. We don't want the mess that the Linux world currently is.

>In my view Sun is under no obligation to relicense ZFS
>to GPL or anything else.

Perhaps you can point out where someone suggested otherwise? Or are you just "typing out loud?"

> If the linux community don't like it - why don't they
> make their own ZFS?

Probably because it's not the ultimate answer to computing as we know it. It's just a very classy file system with some good tools to manage it. Some great ideas in it, no question, but it will only strongly appeal to a select group of business cases.

On the other hand, I would hope that most of Sun's customers would give it a fair shake and use it for new installs of large-scale storage, since it is there for the taking. Really, they'd be silly not to if the real-world reliability and performance are up to par with the feature set (and I suspect that they are).

Probably because it's not the ultimate answer to computing as we know it. It's just a very classy file system with some good tools to manage it. Some great ideas in it, no question, but it will only strongly appeal to a select group of business cases.

There's a lot of truth in that. Specifically, especially today, while ZFS has many very nice features, it will only migrate up to the users via some gaudy flashing icons on the desktop and some kind of user education campaign.

To me the big value of ZFS for end users is simply storage management. As their disks get bigger and bigger, they can easily add more storage to the pool and just grow it endlessly and "never" run out of disk space, never have to move a file again, never have to copy partitions. Just keep slapping drives willy nilly onto the system, and it gets bigger -- just like RAM.

The fact that on top of that you get all of the other wonderful services of ZFS is just gravy.

No one is going to swap OS's solely for a file system.

But it sure makes anyone who happens to be running Sol 10 already a good reason to upgrade, even without ZFS root.

If you know somebody that has a JumpStart server with Solaris 10 6/06 installed yes, otherwise no.

Good thing they don't work for me, I guess. I'd have kicked their ass for using alpha/beta grade software in production environments.
Ah now you are getting personal or kinky?? You wouldn't have a big enough boot. I guess you are still using Solaris 8 or earlier?

The quality of the code before it was release

Let's let Sun be the judge of that. It just hit release, didn't it... oh, yes it did.
Many within Sun consider Solaris Express as a "Release" , and they do not consider it alpha/beta. It comes down to the level of support you require.

I mean I would only use Solaris in a production server. No other taskset.

So you would not use Solaris for development or test environments? If so Why???

And this is different than the Linux zealots modding up comments they like how?