Five years of Sun software under Oracle: Were the critics right? Back in 2010, critics worried that Sun Microsystems' software portfolio would wither on the vine once Oracle got its hands on it. Five years on and the worst fears have proven baseless, yet former Sun diehards have had plenty to be disappointed about since Larry Ellison & Co gobbled the former server heavyweight. Although Sun …
This post has been deleted by its author
He said that he loves Linux, but it is for low end. Solaris is high end.
Of you look at oracle largest and fastest servers, sparc m6 and the new m7, they exclusively run Solaris. These supercluster servers have many sockets, and as we all know, Linux does not scale well beyond 8-sockets. So how could oracle offer Linux on 32-socket servers? That's not possible. Linux is good for small servers, but you have never found Linux on large servers. I invite you to post links to larger than 8-socket Linux servers.
IBM p795 can run Linux,but it is a unix server, and nobody runs Linux on such a server because of bad performance. Hp Big Tux runs Linux, on their unix integrity server, with 40% CPU utilization under full load! which is bad. SGI altix and UV2000 servers are clusters, and exclusively used for HPC number crunching workloads. ScaleMP servers are clusters too.
Linux is good for small servers, but you have never found Linux on large servers
This magazine runs an annual article highlighting the world's top 500 supercomputers, many of which run Linux. As of Nov 2014, the world's top ten supercomputers are all Linux based.
> This magazine runs an annual article highlighting the world's top 500 supercomputers, many of which run Linux
An none of these systems are running single instance of the Linux.
Even kits with hundredths of CPUs are running multiple Linux instances (each one with own kernel etc) as partitioned environment.
Even kits with hundredths of CPUs are running multiple Linux instances (each one with own kernel etc) as partitioned environment.
However many instances there are, the OS of choice is Linux, not anything else. It is strange that Linux dominates the very biggest (top 500) and the very smallest (eg. Pi). Perhaps due to licensing issues as much as anything.
@MadMike,
SGI Altix UV is not a cluster, (distributed memory), but is a ccNUMA shared memory platform. ScaleMP the same, although without HW aided cache coherency lookups.
Old Sun E-series, were true SMPs, i.e. all CPUs connected by a crossbar switch.
Almost all modern day multi-socket servers are ccNUMA platforms.
I don't think Red Hat SUSE, or Canonical support Sparc at all, just x86, Power, ARM, System Z (IBM mainframe), and Itanium (now dropped). Since Oracle Linux is just a clone of Red Hat, they inherit Red Hat's support matrix. Yes Linux will run on Sparc, but I'm not aware of anyone offering commercial support for it today.
Sparc has become a really niche architecture. Let's face it, everything outside of x86 and ARM is niche these days, and your third party software options are pretty limited once you stray outside those two. Power still gets some third party love and attention, but there's not much new market growth to attract vendors to anything else. Even Intel with all their money can't save Itanium.
So why do Oracle not support Linux on their own Sparc hardware? Splitting a small market across two operating systems fragments it too much. Since they have legacy Solaris customers they need to support, they have to support those customers anyway, so that means Solaris for Sparc. It's much the same reason why HP needs HPUX for Itanium. For Oracle's x86 hardware, the market is for Linux, so that's what they need to support there.
As to what the Oracle salesman may tell you their reasons are, well ... they get paid to be salesmen.
I don't think Red Hat SUSE, or Canonical support Sparc at all, just x86, Power, ARM, System Z (IBM mainframe), and Itanium (now dropped).
SLES is only available for x86-64, Power, and z. I don't think openSUSE supports any others. Don't know what platforms Red Hat is supporting these days, but I suspect you're right - if you want Linux on SPARC, you're looking at rolling your own distribution.
".....Linux does not scale well beyond 8-sockets...." Er, have you heard of this stuff called "grid computing"? Very popular with Linux since 1998, there was even a SUN version called Sun Grid Engine but that was a piece of SUN software that didn't survive Larry's stripping of the SUN carcass (http://en.wikipedia.org/wiki/Oracle_Grid_Engine). Mind you, Beowulf is a bit old hat for today's whippersnappers. But today's Red Hat's vanilla clustering scales to 16 nodes, and I can remember building RHEL 5 instances across 16 Itanium2 sockets (and those were dual-cores) on hp Superdomes back in the day and then clustering those instances up into four-way clusters - absolutely trounced SPARC Slowaris in billing system trials. Maybe you need to update your Linux knowledge.
"Linux does not scale well beyond 8-sockets...." Er, have you heard of this stuff called "grid computing"? "
Yes - it's a way of working around the fact the Linux doesn't scale very well beyond 8 sockets by tying together lots of discrete systems...It doesn't fix the fact that Linux doesn't scale very well.
"It doesn't fix the fact that Linux doesn't scale very well"
Any links supporting this claim? Are NASA aware of it, since they have been building pretty big single image supercomputers for quite a while.
AMOF, the fastest and largest SSI (single system image) computing beasts have been running GNU/Linux for a long time now. Not sure which one holds the record right now, one notable example was Columbia (named after the crashed Columbia Shuttle crew). Here's one back from 2007, don't have any fresher examples, sorry.
Here is also an excerpt from arch/Kconfig:
"config NR_CPUS
int "Maximum number of CPUs" if SMP && !MAXSMP
range 2 8 if SMP && X86_32 && !X86_BIGSMP
range 2 512 if SMP && !MAXSMP
default "1" if !SMP
default "4096" if MAXSMP
"
I.e., default number of logical cores is 4096 when MAXSMP is enabled.
"...... it's a way of working around the fact the Linux doesn't scale very well beyond 8 sockets by tying together lots of discrete systems...." No, it's a way of taking advantage of cheap 2-way x86 servers to build a massive solution for a fraction of what it would cost to do the same on SPARC. But, you keep on living with your head in the sand if you like, Sunshiner.
"Ha, ha, ha, Slowaris wasn't funny five or ten years ago when this individual started using it....." Sorry, I can't take any credit for the "Slowaris" moniker, that came from SUN's own customers years ago. Indeed, this definition was posted in 2001 - http://everything2.com/title/Slowaris. You can go stick the term into Google and you'll see hundreds of hits for its useage (Slashdot currently being top). I can only surmise that it has been a long, long, loooooong time since you worked in any enterprise environment.
I was disappointed that they missed something I think give an insight into another reason Oracle wanted Sun. Before the buyout Oracle were saying exactly the same thing as ASF were about opening up the specification processes. Java is essential to just about every aspect of their business, and they'd been buying a lot of companies and tech based on Java (e.g. JRockit, the JVM most people considered the most performant at time.)
Jump forward a couple a short while and suddenly they no longer felt that opening of the process was no longer needed. A huge shame, Harmony had so much to offer and potential.
Java under Oracle had a very shaky start but seems a lot more on track now, still plenty of wrong directions being followed IMHO though. I do wish however someone else bought Sun, I'd have preferred Google or even IBM.
I must admit that 5 years ago I was quite worried about where Java was going under Oracle ownership but all things considered I've been pleasantly surprised. I'm not sure I agree with the article about the rate of releases of Java being better under Sun than Oracle. From my perspective Java was stagnating under Sun, it really felt like they had pushed as hard as they could to get to Java 6 and were exhausted.
From what I understand of it the core Java libraries were a bit of a mess when Oracle took them over and a lot has been re-written. Certainly the JVM feels a lot snappier than it did when Sun was running the show. As for openness, yes Sun started the process of open sourcing Java but realistically it was never going to happen. At least under Oracle we now have a decent open source version (that's actually the reference implementation which is even better).
If I could have one wish for Java it would be that Oracle put more concrete-code-meat on the API bones of what they release. Logging is the perfect example - half a dozen implementations of basically the same thing all of which have their own issues. Then there's XML, JSF, etc, etc.
One fear I do have is that Oracle will kill or cripple GlassFish. It looks like there will be a GlassFish 5 covering Java EE 8 but I wouldn't put money on there being a GlassFish 6.
We used to have over 200 Solaris servers in our (University) business environment. We now have zero.
Everything is now Windows (SPIT) or RHEL Linux. It's all way less of a problem to administer and much cheaper CPU for CPU than anything out of Oracle, no matter what the Larry fanbois on the dedicated we love Larry side of the line might claim.
We are even going with a Cisco/ENC VBlock for our production systems this spring. Did I say it's LESS THAN HALF the cost of an Exa-whatsit set up? RHEL and Virtual disks along with VMWare does it for us. No need for Larry-Lock-In.
Sun screwed the pooch and lost it all to Larry. As a paying customer, we prefer other vendors now. Oracle is indeed a toxic environment. Never again.
Solaris is still alive and well in the financial and telecoms sector. However, I will concede that they aren't buying as much Solaris iron as they used to, Oracle's support has skyrocketed. And sadly, price hikes have meant that not all sold Solaris boxes are using SPARC, with a sizeable chunk using crappy x86 instead.
But Solaris being dead? I don't think so.
It is mostly dead for us and our customers.
With half the money, we can do so much more with RHEL, Jboss, mysql, mongo, etc, or go the IBM route, AIX, DB2, websphere, also cheaper (but not half the cost).
Only the systems that were made for solaris and would be to expensive to test (as they are very complex) are still on Solaris/Oracle, or hosts.
Most of us knew that it was a way better proposition to use linux.. long ago, but as we had to sell the idea, we kept selling Aix/UX/solaris/NT
> We used to have over 200 Solaris servers in our (University) business environment. We now have zero.
As same as you can find many example when may years ago was none Solarises in some environments and now they are or number of Solarises is climbing when in the same time number of Linuxes is flat or is going down.
Fact is that Oracle earns from paid Solaris services now much more than at any time when Sun was around.
Solaris developers teams grow so much that inside Solaris department now is few kernel projects with more developers that in last days of the Sun been working on the whole Solaris.
Problem with estimation of Solaris market is that most of the Solarises are working in internal networks and all these systems are not countable by any public services like http://www.netmarketshare.com/os-market-share.aspx?qprid=9
What a bunch of garbage. Solaris especially on SPARC is nearly dead just like virtually every other proprietary Unix and all you have to do is compare sales numbers from the companies themselves to five and ten years ago to see this true. Yes it might hum along in basically maintenance mode for many years to come but its in the old folks home at this point with virtually no chance of major growth again.
I just can't believe the ignorance of so many people on these sites. People who claim proprietary Unix is dead should just look at what computer or phone they are running. Apple's own Mac OS X and even iOS are based on "proprietary" Unix and clearly, Apple is not nearly dead if you saw their recent fiscal numbers.
And if it wasn't for Solaris, being in the industry for the last 20+ years, outsurviving TRU64, VMS, HP-UX and soon AIX, Linux probably wouldn’t be where it is today and we'd all be stuck in hell with windows. Good thing that didn't happen and that Linux and even Solaris is thriving. Competition is good and healthy otherwise theres no reason to improve. How does the main Linux vendors compete? Certainly not on technology.
What I find amazing is that many of the latest "Linux wannabe features" like containers, Dtrace or even ZFS, still aren't fully baked considering the apparent "zillions" working on Linux today. Why is that? Because 90%+ of Linux development is on mobile platform like Android today.
So you say Solaris is proprietary yet its supported across any vendors x86 system and all vendors of SPARC, an IEEE open standard. What does proprietary mean? And anyway, why does it matter if your running a mission critical business, you need support and that’s what matters.
And actually, if you look at Oracles premiere support model, which is just 8% of the HW net price, its actually considerably cheaper than RHEL or SUSE and all the add-ons and extras needed to match.
So, from where I am looking at the enterprise/commercial market, theres still a lot more risk, and more costs implementing Linux today than Solaris, and you can see by all the Linux kernel developers and "bug" fixers in these companies. Why do you think Fortune 1000 companies OPEX budgets have exploded these last 5+ years-because many have drank the coolaid that x86/Linux saves costs! That might have been true 5 or ten years ago, but if you evaluate things today, its no longer the case.
Yes yes I will agree my wording was poor and for the record I am hardly some Linux fanboi. I actually think much less of it than Solaris especially what with Linux having cancer these days (systemd). Still this article pretty much says it all about Solaris under Oracle from Mr Java himself.
James Gosling slams Oracle’s over Solaris
One big fail Mr Ellison
14 Jan 2014
"Four years after Oracle bought Sun, Java founder James Gosling has waded into the outfit’s handling of his former workplace’s key assets.
Writing in Infoworld, Gosling scolded Oracle on its handling of Sun’s products and was particularly nasty about Larry Ellison’s handling of Solaris.
Giving Oracle an F- for the way it treated Solaris, Gosling said that it was now “totally dead” and it was hard to think of anyone actually using it. “Hardware systems from Oracle make no sense at all. I had to convert all my Solaris boxes to Linux, it made me weep.”
http://news.techeye.net/software/james-gosling-slams-oracles-over-solaris
As a contacting Unix admin, I know at least one large Sun customer who is abandoning Solaris due solely to Oracle's spiraling support costs, poor (support) performance and don't-care attitude. They like Solaris and would like to keep it, but when refresh time comes, Solaris is always replaced with Linux. Sad.
I like Solaris. Although it never made the administrators job easy, that was mainly because it is so cutting edge. It torches pretty much every other OS.
> like Solaris. Although it never made the administrators job easy, that was mainly because it is so cutting edge. It torches pretty much every other OS.
So why systemd is trying in so many details simulate what Solaris SMF does? or btrfs is trying to be ZFS like? Why perf is attempt to be Linux DTrace? :)
SMF, ZFS and DTrace are on the marked ~decade and still so long time systemd, btrfs and perf are not even closer to what is available on Solaris :)
"abandoning Solaris due solely to Oracle's spiraling support costs"
Many organisations in finance (which I'm most familiar with in recent times) will still rubber stamp Solaris for additions to existing (legacy) products, but make you jump through a large number of hoops to justify Solaris over RHEL for greenfield projects.
Let's not beat around the bush here.. Java on desktops (not so much on other platforms) is a heap of shit. Anybody who codes still Java applets for web pages needs to be taken out and shot. Basically, it's a slice of the 1990s where much of the functionality can be replaced by quicker, more stable and more secure replacements.
Oracle's products of course heavily rely on Java. Oracle forms is a particularly obsolete slice of twentieth-century technology that still uses it. Except of course for when your particular version of Oracle doesn't work with the latest version of Java which is always fun.
> Let's not beat around the bush here.. Java on desktops (not so much on other platforms) is a heap of shit
IMO 99.99% of all developed java applications not been developed to run on desktop.
The same Boeing 747 is like shit on transporting sofa inside NY Manhattan area ..
"I'm sorry, but... what?"
It's been run through a few iterations of a somewhat less than perfect translate engine - for example English (Scouse) -> Amanfrommarsish ->Hindi -> Latin -> Apache -> Fortran 66 -> Korean -> English (Jamaican).
1.5% error on each translation, translation order randomised.
I've reversed the process, and have come up with the following as the initial utterance with a confidence level of 95%:
A man with a bong on the 7:47 from Shere to Sofia transformed some shit.
Clearly you've never used any of Oracle's "Enterprise" applications. All packed full of ginormous (and extra, extra fugly) Java applets which are amusingly apt to shit the bed entirely when Oracle releases an update to Java. They broke EBS clients entirely when they did the (almost-)forced update to Java 7 a couple years ago.
Oracle is a horrific place to work. The culture shock for Sun employees was massive. Sun had a tendency to over deliver, especially in support where I worked. But Oracle really couldn't care less if customers stay or go. My boss was told to stop providing feedback and just chow down on the corporate doctrine.
So all the good people have left and only the weak stayed. Awful awful company.
Yes, it is sad but true. One of my previous jobs was at a consulting company where pretty much everyone was an ex-Sun employee. All of them left after the Oracle takeover, and they all left for the same reason: once Oracle took over, they were relegated to second-class citizens in the corporate ladder.
Sad, because for what they've told me, Sun was a place where I would've loved working. Alas, it's long dead.
For my money Java, MySQL, OpenOffice and VirtualBox all suffered. Sun machines and Solaris used to be fairly common in places I traveled, but no more.
Oracle is a company bent on serving Oracle/Larry to the exclusion of anything else. It seems to be a culture that values 'winning' regardless of whether or not it diminishes net wealth all round.
I am mystified as to why people stick with Java when it is *clearly* encumbered by Oracle. A proprietary programming language cannot and should not have a future.
I disagree on the MySQL bit. I don't like it much as a database but I have one project which is dependent upon it and there's no doubt in my mind that Oracle's stewardship of MySQL is better by far than MySQL's own where being fast seemed to trump being reliable.
Of course, the big winner out of the takeover has been Postgres with companies like EnterpriseDB picking up lucrative contracts from those fleeing Big Red.
Picking a fight with Google over Java was stupid. Android finally gave Java the mass market of developers it had been craving. And now that market will go wherever Google leads it.
"where being fast seemed to trump being reliable."
But that's exactly the niche MySQL was designed for -- things like web applications where an occasional data failure could be tolerated but high performance was a must. If it sacrifices performance for reliability, it's playing catch up to PostgreSQL, and it has a long way to go.
"I am mystified as to why people stick with Java when it is *clearly* encumbered by Oracle"
So I'm developing say a web application with a Java back-end. How is Oracle 'ecnumbering' any part of the process I go through?
If you are 'mystified', you probably have not used, do not know Java - or failed to use it correctly when you did.
In large corporations the 5 year time frame has not yet given enough time to see the effects of Oracle's stewardship. Many of us are signed up for multi-year subscription deals and it is only as those come to an end that any thought of departure will arise. Once it does it then takes years to plan an execute the migration.
For myself we're just at the start of what will be a long journey away from Oracle WebLogic and onto open source middleware. It has taken until recently for regulators, auditors and high-up decision makes to get comfortable with the idea of using open source application servers for mission critical platforms in highly regulated financial industries. Now that we are happy with this move then Oracle will shortly be receiving the unwelcome news that we are not going to be renewing the multi-million pound contract.
I'm sure we can't be the only ones who are only just getting to this position?
I hated what Oracle did with a lot of the Sun products, including pulling support and even the ability to download a BIOS ROM for old workstations without buying a pricy service contract - but VirtualBox has become (remained?) probably the most useful Oracle software product to me, particularly because it is still accessible to anyone without having to pay a king's ransom, and it works well.
On the other hand, every single time I install a Java JRE update, I am reminded of how much I hate Oracle, since they provide "fake" options to disable various annoying features which, like clockwork, re-enable themselves without notice every single time you do a bugfix update, or even before then. (One of my long-held pet peeves: SW that wants to "pre-load" itself and suck up continuous resources on a computer when it is not being used, simply because it makes the SW look less like a sluggish albatross when you eventually actually have a reason to use it and 80% of it is already sitting there in memory, playing "memory sponge".)
When Sun was for sale I was hoping against hope that IBM would buy it instead of Oracle. I was in corporate IT when Oracle was young, and we knew them as a bunch of, well, tellers of things not intended to be taken as true statements, while at least IBM has amazing engineers if questionable management and strategy. I'm happy and a bit surprised to see that Java and MySQL have survived, since I spent a large part of my career working with them, especially Java.
Just because some Sun people left, including some prominent ones, doesn't mean everyone left behind is a moron... it is insulting to many engineers who have been working on Solaris for many years and delivered (and still do) many interesting features. Then Oracle has been hiring in Solaris Engineering and the team is now bigger than during the Sun days.
Also, to suggest that Solaris is stagnating is pure ignorance (or worse). Under Oracle there's been a continues stream of improvements and new features. The main difference is that all of it is happening behind a closed doors so there is much less (almost nothing) talk about them in OSS communities. Which is a shame. If anything it is Linux still catching up to Solaris (both x86 and SPARC) in many areas. BTRFS is still years away to be in position to compete with ZFS (and see how ZoL is popular), apart from Oracle's Linux there is still no DTrace equivalent on Linux - yes, there is a bunch of other tools which can get the same job done, but usually it is much harder to use them and most importantly still risky in production environments. Solaris Zones (and now Kernel Zones as well) are still ahead of Linux containers. SMF is still much better and mature than SystemD. There is still no Linux equivalent to Boot Environments which are great (fast, reliable OS updates, etc.) nor to FMA, Solaris does fast reboot in a safe way and most Linux distros (commercial) do not have kexec (which is similar), etc.
Having said that in the foreseeable future the commodity market for server OS is down to Linux and Windows, and iit is not clear that it is Windows loosing the battle in enterprise... but if you need additional features then Solaris is often a better choice than Linux, and in most cases it is cheaper. You can run it on the same x86 HW as Linux and support cost is similar to RedHat.
This post has been deleted by its author
Well, sort of if you ask me... On one hand there is indeed quite a lot of enterprise software - on the other though you see more and more home-user machines - especially Windows ones - which do not even have JRE installed and do just fine without it. The fact both Java 7 and 8 runtimes are still somewhat glitchy GUI-wise under Linux doesn't help either.
That's a pretty good comparison, actually. A huge base of installed software and siloed programmers, on top of a language that was long ago surpassed by a number of alternatives. Desperate attempts to stay relevant which are not gaining converts (anybody remember Object COBOL?).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
