OSNews

The cool thing about SPARC is that it's really fast. I couldn't believe the results of a $5000 Sun T1000 server, just a few weeks ago. And it's entry level that's going to be EOL-ed.

Currently they are investing in two areas in Solaris: Storage and Networking.
If the Solaris storage part (ZFS, Lustre/ZFS and Cluster) get's a few more features it's going to fly by NetApp, EMC and the rest. It's almost a complete solution. Regardless of what anyone thinks about ZFS, it's like comparing Windows XP and Snow Leopard. ZFS just works and it works easier than anything else on the market. Storage administration is fun again thanks to ZFS. Sure, you might be able to do the same things with solutions from other vendors, but they either come at a very high cost or they are less than easy to administer.
While the networking part is also getting some serious love, I'm not sure that it's ever going to be competitive with Cisco or Juniper. Linux or the BSDs still aren't at that level. For easy tasks, they work, but if you need firewall, content inspection, routing and QOS, it starts becoming a head-ache. A free head-ache, but a head-ache nevertheless. I can't imagine Solaris being less of a head-ache. It's probable going to be a bit easier to administer in some aspects, but I don't think a general purpose OS can go there.
Everything in Solaris fortunately just works as expected. Furthermore, it's the only UNIX out there still in development (AIX or HP-UX don't seem to get any love lately), and it's that good that it's actually worth giving up RHEL for it.

That's not really saying much. At all.

As we're in the business of trolling, I return a favor by saying:
Luckily for RedHat, not many people share your sentiment.

- Gilboa

There are custom platforms / appliances built around Linux or xBSD that'll do pretty much whatever you want.

Here are but 2 examples:

Untangle Open Source Network Gateway
http://www.untangle.com/home

Vyatta - Routing, BGP, VPN forwarding, virtualizable

http://www.vyatta.com/products/product_comparison.php

Cisco has new switches running Linux and they're rumored to be doing some routers in the future that are Linux based emulating catOS.

Juniper's routers are based on a customized FreeBSD.

You see, the networking capability of these OSes is already there.

I think AIX / HP-UX must EOL'd in favor of Solaris. It is UNIX, it comes with a free desktp version and more importantly there are people who love it and maintain it. AIX/HP-UX are closed as charged.

Juniper (and many other vendors) use FreeBSD as a base for their systems and are contributing a lot of code back, check this for example:
http://www.internetnews.com/dev-news/article.php/3835746

It seems you have not spent much time with Cisco. Easy configuration is not a strong point. Firewalling, routing and QOS is all easier to set up on Linux/BSD than on Cisco. Or at least on OpenBSD, can't say I'm a fan of iptables.
It's also a bonus not having to deal with the awful CLI interface of IOS (it almost makes edlin look user friendly).

Note: Juniper routers run a custom version of FreeBSD.



And MacOS X is also certified UNIX, and still under heavy development.

Edited 2009-09-11 17:44 UTC

From Cisco's giant Nexus 7000 10-Gig switches, Juniper megaswitches, F5's BIG-IP/LTM, to consumer grade wireless access points, I don't know of a single network vendor that uses Solaris (or OpenSolaris) as a base operating system. Cisco, Juniper, F5, all the major networking companies use Linux or FreeBSD (or the other BSDs).

Edited 2009-09-11 22:29 UTC

Thats what I admire about Larry - he has balls of steel; willing to make a move and not pussy foot around; thats a bit of a contrast when compared the namby-pamby directionless leadership which occurred under McNealy, Jonathan and Zander - Zander, the only one that can take a heathy company and kill it over night just by his mere presence. I swear the guy could make flowers die just walking past them.



If anything is going to happen it will be Larry focusing on the corporate desktop and server - if 100% of his effort is spent on the corporate desktop and server, everything else will fall into place. The problem is people who attack Microsoft attack form the wrong angle. They assume that the key to winning is getting the home user on board. The home user is merely a follower to what ever is in place at the place they work.

OpenSolaris is a wonderful system that is neglected - hopefully with HAL being retired in favour of DeviceKit (and friends), NWAM Phase 1 finally getting merged which will deliver a great networking experience for the end user, improved power management as more of the system is made tickless and hopefully a greater focus on reducing the memory hogging that ZFS does. If they address those and GNOME 3.0 really takes off - it'll be a great operating system for the Toshiba laptops that are being sold right now with OpenSolaris preloaded on them.

There have been a lot of suspiciously similar advertisements over the cours of the last ten years or so.

I knew things were bad, but I didn't know they were this bad, not having been in the 'Sun loop' for years:

http://www.internetnews.com/bus-news/article.php/10791_3832666_2

The big customers are vitally important to Sun as that is what keeps the revenue relatively high. It's impossible to believe Oracle can do much, especially considering the takeover still hasn't happened yet.

"It's impossible to believe Oracle can do much,"

I believe Oracle can do much. Hence, you are proven wrong.

For instance, I believe one of the first things Oracle will do, is to change the pricing on Oracle Database so to not punish multi core SPARC. I believe SPARC will be favoured over the legacy Power6 and the other CPUs.

Since when "I believe" has been an acceptable scientific proof of anything?

Not that I disagree with what you are saying. But seriously... you need to provide facts and references when disproving somebody else's claim, not simply stating what you believe as irrefutable proof for anything.

Wordplay at work that went over your head.

The claim: it's impossible to believe
The rebuttal: i believe

Hence, the claim is false.

Logic and reasoning over yours. ;-)


The problem with that logic is that it's a pity that I backed my belief up with evidence from two quarters of a pretty clear, steep and marked decline, with an article on the very subject that I quoted from. :-)

As the parent said, you need to cite some facts and references as to why the claim is more likely to be false. Only then can there be any meaningful discussion and that's why there is no response to the OP above. ;-)

The core talent hasn't gone anywhere. I have no idea why people continue to believe this.

didn't you know they all jumped to that other sparc specialist... no wait... they prob sat in their jobs continuing to work and be paid.

"didn't you know they all jumped to that other sparc specialist... no wait... they prob sat in their jobs continuing to work and be paid."

ya its a shame no one else makes a SPARC product... oh wait. http://www.fujitsu.com/global/services/computing/server/sparcenterp...

In fact, Sun liked Fujitsu's chips so much... ah forget it, i don't have time to feed the trolls. (thats not true, i do have the time, but i just don't feel like it today).

Things don't look bright there either and Fujitsu are naturally gravitating to where the growth is:

http://www.channelregister.co.uk/2009/03/30/fts_server_strategy/

it was a general sarcastic post. I am aware there is more than 1 SPARC producer.

I was trying to portray that people don't suddenly quit, on mass, when they are in continuous employment. In particular when your specialism is in a niche sector

Probably because there's a lot of pretty strong rumours and firm events to suggest that all is not well, and many are people who have been with the company for a long time?

http://www.theregister.co.uk/2009/04/08/sun_chip_geek_quits/ (this guy definitely qualifies as 'talent')

http://www.theregister.co.uk/2008/10/23/bechtolsheim_still_at_sun/ (which seems to be some sort of 'Buy my company' arrangement)

http://www.theregister.co.uk/2008/03/25/sun_yen_juniper/

They also laid off about 2,000 people in September 2008, a lot of them skilled workers who had been at Sun for some time. Another few thousand went this year, and these were just the latest in a long line of culls over the past few years.

Quite how you can say that there isn't a problem with talent leaving, distractions and morale there I don't know. It's an open 'pretend' secret in many circles and the result of a classic case of being in a market where costs are too high but not being able to maintain what you have when you cut.

Probably because there's a lot of pretty strong rumours and firm events to suggest that all is not well, and many are people who have been with the company for a long time?
...
Quite how you can say that there isn't a problem with talent leaving, distractions and morale there I don't know. It's an open 'pretend' secret in many circles and the result of a classic case of being in a market where costs are too high but not being able to maintain what you have when you cut. "

I'll put it very plainly, I know *many* of the engineers that work on Solaris at Sun and they haven't gone anywhere. The media likes to make big noise about people that leave whenever a company is having problems, but that's all it is: noise.

It's true that Sun has laid off many people, but so did a lot of other companies over the last two years (hello, recession!). Yet, despite that, I can tell you with certainty that Sun has kept its core talent, and none of them really want to leave.

I'm sure you do, but that doesn't paint over the thousands of people who have been laid off over a number of years and the people who are still being laid off recently. So many have gone that you have to wonder who else can go now - and operating costs versus revenue is still high.


If you see lots of smoke then the logic goes that there is a fire somewhere. That logic hasn't changed, especially when a company is having known problems.


It hasn't just been the last two years and I'm afraid it hasn't been because of the recession - the recession has just pushed things over the edge to the point where Sun needed to get bought out quickly:

http://www.channelregister.co.uk/2009/03/31/sun_layoffs_confirmed/ - 2009

http://www.theregister.co.uk/2008/05/01/sun_q3_down/ - 2008

http://www.theregister.co.uk/2007/08/07/sun_to_axe_more_jobs/ - 2007

http://headrush.typepad.com/creating_passionate_users/2006/08/silve... - 2006 - they'd been getting the boot over a period of four years

http://news.cnet.com/Ex-Sun-exec-lambasts-Suns-late-layoffs/2100-10... - they'd axed cut more than 13,000 jobs between 2001 and 2005, but expenses remained high.


That remains to be seen. There's a bit of noise there when a lot of smoke is being created. What Sun needed, years ago, was to get products to market quickly in the market that they have decided they want to be in. The Rock delay debacle, amongst other things, wasn't what they needed and despite all the talk over Niagara single-threaded performance is what people demand.

Despite Oracle's reassurances to existing customers they aren't going to let their cash pile be frittered away on competing with IBM in particular, in a market where they've been losing out for some time.

SEGEDUNUM,
Earlier I complained that you claimed lots of false things without any evidence. For instance you claimed that ZFS need several GB RAM - which is a wrong statement. People have run ZFS on 512MB, and FreeBSD's ZFS implementation had a bug which made ZFS require much RAM. I asked for proof of your claim, but you didnt provide any. Still you continued to claim. That was wrong of you.

In this post, you provide links and you are clear that they are your opinions. You dont state as facts, but as your opinions. You can not say that "ZFS needs several of GB to function" as a declarative fact, when it is not.

Because you provide links and proofs, and you state your opinions, I have no problem at all with your post. Sure, I dont agree with you as I am a SUN supporter, and you strongly dislike SUN - but that is totally ok. You can disagree with me, and you can write whatever bad things you want, about SUN - as long you are clear that it is your opinions and that you provide links. I will never complain on your posts, then.

You can hate SUN or whatever in your posts, it is ok. I will not argue against you. Everyone has right to have an opinion. If you dislike SUN, that is ok.

Just wanted to make this clear to you. I will not attack your posts if you continue like this: opinions and proofs (via web links). If someone attacks your posts, I will defend you. Because everyone has the right to have an opinion.

But if you write false things, you will have a problem with me. Like "Niagara SPARC is slower than Power6" - which is not true. Some benches SPARC is many times faster than Power6. That would be a false statement, that needs to be corrected. If you write "I like Power6 better" - then that is ok. Because it is your personal opinion. Other people wont be misled to believe that Power6 is faster, they understand it is your personal preference.

"I may not agree with you, but I will give my life for your right to express yourself" - some clever guy said so long time ago.

That's mighty big of you, I am not so kind. I don't agree with the reason why segedunum throws Sun under a bus. It stems from his experience with Solaris 9 and his attempts to run Zope. According to segedunum Zope runs far better on Linux than Solaris. He does not articulate why, or what he did to come to this conclusion. I don't buy it. Where is the performance data to back it up, did segedunum run sar and system accounting to gather some data, more than likely not.

This is no different than some of the application owners I deal with at work who say the OS and the hardware are at fault for poor application performance every time. In order to get optimum application performance you have to gather statistics and tune the application, anyone who expects an application to work perfectly out of the box is an idiot! In one of the instances I just mentioned, the default Java parameters were set to 128 MB, the default. We set them to use a GB of memory and the performance issues went away. Adrain Cockroft states "tune the application, not the OS" and this man knows far more about performance tuning than I do. So if the application performs poorly is that the fault of the OS? I don't think so. To say so is nothing more than a cop out and shows a lack of administrative skill.

Segedunum's so-called facts are nothing more than articles and blog posts that support his position, anybody can cherry pick pieces that support their argument. Ask segedunum to point to something he actually likes about Solaris, I am willing to bet he can't. At least shaman, the troll who said that OpenSolaris is "vaporware" could point to several things he liked about Solaris. My position on Linux here is well known, but I can point out things I like about it, just as I can with Windows, AIX, and HP-UX.

If segedunum is still going on about ZFS and memory requirements, he obviously chose to forget the last article I wrote for OSNews where I benchmarked the performance of ZFS against ext3:

http://www.osnews.com/story/19823/A_Solaris_Administrator_Looks_at_...

I use ZFS in production and have no problems with it and have stated so here on numerous occasions.

My personal belief is that segedunum has minimal to no experience with Solaris 10 or OpenSolaris, I can see discussing issues with someone who actually has experience with the product, I think all he does is search for articles and opinions that support his position and posts them as his "facts". I at least use the product and point out what is good and bad about it.

I just think your support is misplaced and if you want to defend segedunum that is your choice, I wouldn't and don't intend to.

It's one of the reasons, but not the only one and it's far bigger than that. A lot of organisations have ditched Solaris and SPARC over the years because of not only poor performance but a complete inability to get popular open source software running properly. Apparently, we're all supposed use some supported and expensive J2EE solution or recompile in Forte or whatever it's called now.

Only now has a response to that been created in OpenSolaris, and it is just under ten years too late and completely inadequate - with two or three years of vapourware press releases that Shamen was right about. Even now, the only definable reference implementation you can call OpenSolaris is Nevada.


It was to do with Python, threading and SMP systems as I recall, and basically the Python developers effectively said that they weren't interested in troubleshooting problems on a proprietary system on expensive hardware they couldn't run locally and didn't care about. I can't say I blame them. That was 2000. I believe I explained this umpteen times elsewhere.


It is coming up to about ten years ago now but as it was, we ran a comparison test system on x86, BSD and Linux (the perfectly sensible thing to do by the way, rather than fannying about as you suggest) to see if it was something else unrelated, and it ran without problems on both and was about three times as fast. Solaris on x86 wasn't an option at the time. Rather than troubleshoot a complex problem on a system Sun wouldn't help out with, guess what was decided? Why the hell should I or anyone else help Sun keep a customer, and why should you be arrogant enough to think that anyone will?

My performance data would be useless anyway because it's a drop in the ocean. What we want to know is if other people feel the same way. If you want some brutal performance data as to whether that is the case, look at Sun's last two quarters in particular:

http://blog.internetnews.com/apatrizio/2009/07/suns-q4-outlook-clou...


Where are your 'cherry picked' pieces to support your 'position' by the way, and what relevance do they have to the subject of the article?


When you see ZFS on a 128MB NAS box, where it should be an ideal fit, give me a call. I'll say no more because we're getting off-topic and having the issues at hand clouded by technical bullshit - as many Sun consultants do before they wave NDAs at you over a non-existant ECC memory problem. ;-)


Alas, we're miles off-topic now and we are going off down an avenue that has nothing whatsoever to do with the article. Like most Sun consultants they fight bottom-line thinking and problems with technical bullshit in an apparent masking attempt, even to the last. When the brown stuff hits the fan and the fan finally burns completely to a crisp I just don't know what you'll do.

I dont agree with you. If SEGEDUNUM thinks Solaris 8 is a piece of shit, let him think so. He probably has some reasons to think so. Maybe some old benchmarks in the year 2000, ten years ago.

If he provides links and such, supporting his view, then it shows that he has not made up his claims, he is not dreaming. There are people sharing his opinions. Hence, their opinions are valid and are not to be dismissed. Then SUN should use their opinion as a feedback to make Solaris better.

As long as nobody claims false facts, I have no problems at all. If they make clear that it is their opinions, they can write whatever they want. I think.




Segedunum,
Regarding the RAM requirements for ZFS, you say that 128 MB RAM is more appropriate than 512 MB. That is fine, that is your opinion. As long as you dont claim that ZFS NEEDS several GB to function - which it dont.

However, I dont agree with your opinion. My opinion differs from yours. ZFS is targeted for Enterprise use, and Enterprise servers with lots of GB RAM. If you target for 128MB RAM, then you have to make some ugly programming that messes up the ZFS code, you have to make special cases - "if you have low RAM, then use this special algorithm, if you have 512MB RAM, then use this more general, neater algorithm". If you have 512MB RAM as a target, then you can program more freely, and the code will be neater and there will be less special cases and less bugs. In my opinion, a server Enterprise filesystem should target at least 1GB RAM and not less. If you only have 128 MB RAM in your computers, then server enterprise file systems are not for you, obviously.

The good ones have already left. "

No, they haven't. If you were involved in the OpenSolaris community, this would be obvious.

What OpenSolaris 'community' would this be, exactly? The only reference implementation you are allowed to call 'OpenSolaris' is Sun's and that's Nevada. It's not exactly a community by any stretch. It seems that many Solaris engineers are quite right to be worried about the future though:

http://practical-tech.com/operating-system/the-state-of-oracle-linu...



The software side of the business would seem even more certain than the hardware side, because software is Oracle's business. If they want to keep the hardware business as they say they want to then they will neeed very deep pockets indeed, as well as talent.

What OpenSolaris 'community' would this be, exactly? The only reference implementation you are allowed to call 'OpenSolaris' is Sun's and that's Nevada. It's not exactly a community by any stretch. "

Sorry, but unless you've ever been to the OpenSolaris Developer's Conferences (I have), or been to JavaOne, or been part of the OpenSolaris community, then you really aren't qualified to say what is and what is not the community. Since I've done all of that, I am.

I know many of the engineers at Sun, you do not.

I've been to JavaOne, Community One, I'm guessing you have not.

I have been part of the OpenSolaris community (which you claim doesn't exist) since early 2006, and you have not.

So far, you haven't done anything but quote rumours, and made wild accusations about Sun's core Solaris talent leaving. None of which are true. So really, isn't it time for you to admit you were wrong and move on?

Are you implying that Sun is the same as Microsoft? The never ending saga of segedunum posting and commenting on the impending demise of all things Sun and Solaris is an old story, one that should come to an end.

No.

I was implying that it might be the same for some people.

Because, as you can see above, usually one can read such quality comments when the discussion is about Microsoft.

Different companies, but what does it matter when you are on a crusade against this mythical "proprietary software".

Freetards.

Freetards, I like that. Well put.