Solaris 12 disappears from Oracle's roadmap In late 2016, The Register received credible-but-ultimately-unverifiable reports that Oracle was scaling back Solaris development, perhaps with significant sackings. We chose not to publish because Oracle denied the specific allegations we'd received. We’re aware of rumours about the demise of Solaris. Our best efforts have …
"What the roadmap does tell us is that the new OS will debut in 2017"
Or not.
The rather large box fades in and splays itself comfortably over 2017 and most of 2018. Generally in "arm-wavy roadmap" -speak, that means "Sure, 2017... or so."
Given that they're making a seemingly abrupt shift from Solaris 12 being the target development base for new features to Solaris 11, it doesn't seem likely that it's going to be ready any time soon.
It certainly does look vague.
I'm wondering though, is there any really big deal? I'm not entirely sure how one would define where an OS ends and applications start, and whether a lack of development in the OS is a major deal.
The Linux kernel gets updated a lot, and that's pretty cool if you're wanting the latest and greatest kernel features for x64 chips and the eradication of vulnerabilities, etc. However, they're very careful to not break user-land (I can imagine that doing so is good fun - Linus baiting as a recreational activity). But Solaris's kernel, on SPARC, just how much updating is actually required to keep it viable and useful? I can't see Oracle not supporting it on newer SPARC silicon, or failing to fix identified vulnerabilities.
I can see it becoming a problem though if some major user-land package becomes desirable, and the authors of it have gone and made something hideous like systemd a dependency. Solaris doesn't have systemd. Oracle would have to do quite a lot of porting then to accommodate such a package. There's a big trend these days for containers, things like that, but Solaris already does pretty well in that direction, and Sun were amongst the pioneers of quite a lot of that stuff anyway.
But without that kind of issue, most user land stuff should run just fine on Linux or Solaris with no real compilation problems. Solaris can even run Linux binaries without recompilation if required.
Is there perhaps some parallel between Solaris and Windows 7? Windows 7 was, arguably, perfect and pretty much complete, perhaps could have benefited from some under the hood tweaks, etc. Instead MS took it upon themselves to dispose of the entire existing software base (well, demote it to a 2nd class status with a not-shown-by-default desktop) when they did Windows 8. Then 8.1 drew back from that a little bit, 10 more so. The proper, boring, and almost no effort thing for them to have done was simply to tweak 7's innards (for example, 8, 8.1 and 10 have some good improvements to certain aspects of the kernel) but otherwise left well alone.
So perhaps Solaris is in that same place - there's literally not much to be done to keep it rolling along just nicely.
I'm not entirely sure how one would define where an OS ends and applications start
This is actually a pretty important concept. If you don’t have a crystal-clear boundary between the OS and applications, you don’t really have an OS, and application compatibility across releases is a vague fantasy rather than reality. This was the rap against Linux since practically forever: that the OS interface guide was called "the kernel source code."
“and whether a lack of development in the OS is a major deal.”
Spoiler: it is. OSes keep evolving, either with outright new developments (think DTrace in the Solaris world), or co-opting and integrating features that had previously been bolted on outside the OS (virtualized compute, virtualized networking, virtualized storage come to mind). Security continues to be ripe for development.
An example of how lack of development can bite you: When Oracle bought Ksplice in 2011, it was a unique technology, allowing Linux sysadmins to apply kernel fixes without rebooting. Oracle has been talking about how they've been working to get similar functionality into Solaris ever since then, with Solaris 12 as the target release, which would be cool, but at this point it would now be catchup rather than a leadership feature.
So yeah, if this move means functionality like this is seriously delayed or even dropped from Solaris, it will be a major deal. Ditto for their work on integrating Docker. Solaris Containers/Zones is already great, but but adding Docker integration would have been extraordinarily useful.
"Solaris doesn't have systemd. Oracle would have to do quite a lot of porting then to accommodate such a package."
The Solaris equivalent to systemd is SMF, which came in with Solaris 10 in 2005. It was considered to be heretical by Unix traditionalists, but it was a major leap forward in service management and reliability. It’s another great example of “development in the OS being a major deal”.
There’s a lot more to OS development than just keeping existing userland programs running, or even accommodating new hardware. It’s the underlying plumbing, carefully hidden away from the applications (or presented as new services with stable interfaces), where innovation can really pay off. Killing off Solaris 12, which has been in development for over five years, and trying to shoehorn that work back into Solaris 11, is going to make it even harder for Solaris to stay current with OS trends than it had been.
What I’ll be looking for is signs of any of the advanced work that was going into Solaris 12 (and there’s more than what we’ve been talking about here) suddenly making its way exclusively into Oracle Linux.
I can't help but agree. It's a pity if Solaris becomes moribund. I still occasionally use it, but I used to use it lots (weirdly, on embedded systems as well as workstations).
I think you're right - if they start pushing things out for Linux ahead of Solaris, the writing is on the wall.
Anonymous> Wim Coekaerts has suddenly started actively tweeting about Linux on SPARC.
Linux on SPARC has been coming since 2010, at least… with Oracle hiring people to re-write Oracle VM Server for SPARC documentation to remove the exclusive Solaris support in 2015q4.
http://netmgt.blogspot.com/2015/10/sparc-oracle-linux-coming-soon.html
The real question is whether Solaris 11 is the Control and Service Domains or whether Oracle Linux is...
Linux on SPARC has been coming since 2010, at least… with Oracle hiring people to re-write Oracle VM Server for SPARC documentation to remove the exclusive Solaris support in 2015q4.
That feels more like "Linux on SPARC has been talked about since 2010, at least, with something actually starting to happen in 2015q4."
They publicly announced Exadata SPARC systems at the last OOW, and word was some factions were trying to get something out the door at least a couple of years before that.
What's new now, as in "this month", is enthusiastic talk about it from the guy in charge of Linux at Oracle, who sounds like he's suddenly got more shiny toys to play with now that he's back from MSFT.
The control / service domains bit is interesting, but my question is whether they're ever going to run the storage cells on SPARC.
Back in the noughties, before Oracle bought Sun, Sun were hemorrhaging customers as Sparc & Solaris were seen as overly expensive products. Even the Sun resellers said that they were seeing many of their customers switch to x86 & linux. Anyone who's still running Solaris/Sparc should have formed an exit strategy by now.
The fact that Oracle's now being very vague about future on-prem Solaris & Sparc is not exactly unexpected.
The fat lady isn't singing yet, but I think she's starting to warm up.
The multi-decade record of SPARC and Solaris platform development and delivery continues with new innovations going forward. Engineering focus on SPARC and Solaris is being continuously applied to leadership in security, scalability, and enterprise reliability for mission critical computing for key customer adoption opportunities in the Cloud and on-premises.
Future features and functionality in Solaris will continue to be delivered through dot releases instead of more disruptive major releases. This addresses customer requirements for an agile and smooth transition path between versions, while providing incremental innovation with assured investment protection. We are amending the Support lifespan for Solaris 11, to extend it considerably beyond any reasonable expected lifetime of use, through at least 2031 and 2034 for Premier and Extended Support, respectively.
See page 37: http://www.oracle.com/us/support/library/lifetime-support-hardware-301321.pdf
Linked off of this page: http://www.oracle.com/us/support/lifetime-support/index.html
"Solaris 11 follows a Continuous Delivery model, where new functionality is delivered as updates to the existing release; upgrades are not required to gain access to new features and capabilities. As a result, Support dates are evaluated for update annually, and will be provided through at least the dates above."
If any of Oracle's customers require an email communication from an engineering executive in summary of the above, Oracle are happy to do so.
"Solaris 11 follows a Continuous Delivery model, where new functionality is delivered as updates to the existing release; upgrades are not required to gain access to new features and capabilities. As a result, Support dates are evaluated for update annually, and will be provided through at least the dates above."
This makes sense. Big bang releases are pointless and cause customers pain, plus OS migrations every X years are a complete pain in the arse. Incremental updates make sense imho.
That said, what is telling is the lifespan support statement. If they were to carry on development for Solaris then point releases would also have their support dates amended in perpetuity. Ultimately, all vendors want you to consume cloud services (preferably theirs if they have one) rather than roll your own, this consumption-led model also appeals to bean counters across the globe.
From a purely personal perspective, another 15 years of supporting Solaris and migrations to Cloud or otherwise suits me just fine, its just a little sad that finally it looks like Solaris is finally entering its sunset years.
Added on top is the fact that if you come out with release 12 many customers wont migrate whereas they will from 11.x to 11.y, even though its the same code.
The same thing applies even more with the big SIs who want several zillion to do an 11 to 12 but only one zillion for 11.x to 11.y because obviously its easier even though of course its the same code.
What determines its a dot release or a jump to the next number is purely marketing.
Remember that Oracle's first release was 2.0, since "no one would take a 1.0 release"
"What determines its a dot release or a jump to the next number is purely marketing."
Yeah, that's crap. At least for Solaris; I do understand in other OSes, some plucky Finn occasionally wakes up one morning and says "increment it!" That's what you meant by "purely marketing", I guess?
In Solaris-land, there's a solid taxonomy that's been in place for literally decades as to what can go in major, minor and dot-dot releases. The only thing you could call "purely marketing" was the decision years ago to not use the major number in the marketing name, starting with Solaris [2.]7.
Clearly haven't looked at any price, price/performance or TCO comparisons. Todays SPARC M7 and SPARC S7 systems offer *significantly* lower TCO than x86 or Power systems.
http://research.esg-global.com/downloadfile/OracleSPARCM7?filename=ESG%20Lab%20Review%20-%20Oracle%20M7%20Perf%20-%20Dec%202015.pdf&displayFilename=ESG%20Lab%20Review%20-%20Oracle%20M7%20Perf%20-%20Dec%202015.pdf
http://www.oracle.com/us/products/servers-storage/edison-sparc-s7-wp-3395977.pdf
http://www.oracle.com/us/solutions/oos/database/edison-oracle-software-in-silicon-2971285.pdf
This post has been deleted by its author
Not sure if that is still true.
If you are doing large database transactions; with fair bits of OLP or in-memory processing desirable *and* your query load is intense enough to be well in the double digit GB/Second range - then modern architectures, like the M7, provide you with a lot of bang for your buck. Investment not dissimilar; but lower power; higher peaks and less expensive inter-machine due to less kit.
And if you look at the TCO over 3 to 7 years - assuming a tyipical 2-3 hours/24hours at those type of intense loads - you are easily talking half to a quarter of the cost.
If you are doing that sort of analytic queries f24x7 - then the 4x difference in queries throughput mean that the difference gets even more extreme; half the kit and 20% of the power consumption.
* i.e. the linux equivalent of a 2x18+ core E5/2699 and 1/2 of a Tb of ram.
I can't see this going down well at all.
It's likely to be the customers like that who asked for it. S11 takeup is steady, but slow, and neither customers nor ISVs will want the disruption of yet another major release yet, with all the update & recertification issues that it would entail. Add that to the enthusiasm for cloud, and there doesn't seem any more need for a flashy new Solaris 12 than there was for Windows 10. At least Oracle seem to listen a bit better than Microsoft...
This post has been deleted by its author
Like moths to a flame embracing outsourcing, same with Mobile system management back ends (used to be all Solaris and Sun).
Remember when a high proportion of Web servers ran Solaris on Sun HW.
Follow the money. What areas (A) give Oracle best ROI and (B) most Gross repeat sales, (C) Profit?
How many more years of Sparc silicon development will there be?
I have an acquaintance who worked for many years in Solaris support in the UK, both at Sun and Oracle. Last summer he was repeatedly offered a transfer to the Linux team. I told him to take the hint, and he noted that the Solaris team headcount was constantly being reduced. He moved, even though it meant a pay cut.
In ISP/Telco land, where I live, ten years ago *everything* ran on Solaris. Now *nothing* does.
Pity, it's got lots of neat stuff, but of course nothing that can't be ported to Linux.
Must not be running Oracle DB, must be a small shop, your datacenter must be huge and/or only 10% utilized, must run more open source SW than off the shelf, must not use VMware, your business must make more money than it knows what to do with, your endusers-stakeholders-clients must be satisfied with less than or just adequate performance.
if you are large enough to have SPARC, Power, Itanium, still have PA RISC yet you are moving to hyperdense x86_64 your employer is on a roadmap to going out of business. Perfect example of the Right hand not knowing what the Left hand is doing.
I'll go out on a limb and state if you still have PA RISC along with the other platforms, the SPARC, Power and Itanium are probably 2 if not 3 generations old so even a Raspberry Pi would appear more powerful.
You know, "this is the last version of Solaris, everything else will be an update to the existing version..."
People really seem to have forgotten what version numbers are actually FOR these days - they're to communicate that something significant has changed and this new version might break stuff...
"People really seem to have forgotten what version numbers are actually FOR these days - they're to communicate that something significant has changed and this new version might break stuff"
The other side of that coin is that if version numbers aren't being changed they've decided to stop breaking stuff. We can always hope...
"The other side of that coin is that if version numbers aren't being changed they've decided to stop breaking stuff."
By gum, you've solved it! Everyone: stop breaking stuff! And ponies for everyone!
Seriously, Solaris has led the way in not breaking stuff unnecessarily, not obfuscating when there are potential breaking-stuff moments, and standing behind developers when stuff (inevitably) breaks. The last time there was a major compatibility break in Solaris is when they moved some networking functionality from libc to the kernel back around Solaris 2.5 or 2.6 (it's all fading into the sands of time) -- and that only would bite you if you ignored a decade of people yelling at you not to use static linking.
This is what I was thinking. HP led the way with this by rebranding HP-UX 11.11 as HP-UX 11i. Making it the 'permanent' version of HP-UX way back in 2000. There were nominally some follow on releases like 11.20, but they were known as 11i version 1.5 and so forth. Then they got to 11.31 (11i v3) and quit doing even that, simply updating it every 6-12 months (last was 11i v3 update 15 in March, so 16 is probably coming soon)
On the other extreme you have Firefox and Chrome, which have a "major" release every couple months...
OS X did versioning on the minor numbers, i.e. OS X 10.1, 10.2 and so forth. Maybe they'll do it for Solaris and we'll see Solaris 11.1, 11.2 etc. or maybe they'll use the HP-UX 11i model.
Either way, I suspect Solaris won't last long enough to make it to the 10th update if they are done yearly....but it will probably outlive HP-UX.
“Future features and functionality in Solaris will continue to be delivered through dot releases instead of more disruptive major releases.”
"Solaris 11 follows a Continuous Delivery model…”
“It's likely to be the customers like that who asked for [no Solaris 12]. S11 takeup is steady, but slow, and neither customers nor ISVs will want the disruption of yet another major release yet”
I’ve seen this movie before.
After Solaris 10, word came from the executive suite that Solaris 11 would not be coming out for… a while. Possibly ever. Can’t spend all that money on ISV adoption, customers are ascairt, and so on. Let’s just keep on adding things to Solaris 10.
Customers would hear the “Solaris 10 forevarr” message (aka “continuous non-disruptive delivery”) and say, “that sounds great! So when do we get feature X in Solaris 10?”
“Oh, well…” —kicking of ground with toes ensues— “…that requires feature Y, which would require too much change for a dot-dot release…” (remember, as someone already pointed out, “Solaris 10” is really SunOS 5.10, so “continuous delivery” would come through dot-dot releases.)
Customers in general do not want anything to change, ever… except for, of course, the new things. Can we have all the new things, please? But don’t change anything!
When Oracle took over, they saw what was in Solaris 11 and never coming in Solaris 10, and listened to customers who wanted those things, and said, “ship that sucker.” Which took another almost two years, for a total of almost seven years between 10 and 11 by then.
(Oracle had a novel way of solving the adoption costs, though, by not spending anything on customer or ISV adoption. Which might explain the “takeup is steady, but slow” —after FIVE YEARS— part.)
The good news is that a big reason why “continuous delivery” was a non-starter for Solaris 10 was that all the features needed to make it work at all, most notably IPS, were in Solaris 11. So it will definitely be easier to do this than it was 10 years ago. The questions are how much Oracle is willing to spend on backporting how many features from Solaris 12, and at what point do you start fudging the line between what is and isn’t a dot release?
The rain gage to me, and what I’d be asking as a customer, is, “When do you expect to continuously deliverate zero-downtime patching?” since that is A) on the top of the list of Solaris 12 features they’ve been talking about, and B) something that if it makes it into Solaris 11 would be strong proof that “Continuous Delivery” is doable, even for something requiring what would seem to be major changes to the kernel.
Oracle are only really interested in big repeatable sales and minimising costs, it's ALL about the bottom line; this is not nice, but can be smart. Where this goes wrong is when cost minimisation means that support can't function efficiently because of outsourcing to no experience staff in cheaper countries, making experienced people redundant, and poor issue routing.
Customers who expect Oracle to accept piddling revenue or unreasonable hassle must be stupid (I heard about plenty who Oracle dropped); even some established small business take this attitude to make better use of their time and sanity, so that they can grow.
Not surprised, we are paying an arm and leg for the annual subscription license.
The biggest advantage for us is the ZFS file system. This is not rock solid on linux, so I would like to see if Oracle are going to transfer their technology code to Linux kernel or whether we should be looking at freebsd.
Like all great OS'es - when the IQ level of the people using it has diluted to sub 80 the needs for functionally rich software isn't required.
Off to Windows 10 you go - Linux is just a step down on your way.
Your jobs will soon be done by robots anyway.
Famous Quote "Sun is alone too but IT still shines"
SGI made the same move with IRIX 6.5: it was the last major release of IRIX, with only updates thereafter. 6.5.30 was delivered in 2006, 8 years after 6.5. 6.5.30 was vastly different to 6.5.
There is likely many issues here ... the expense and length of the Solaris development model being one of them. Solaris 10 was 2006, Solaris 11 2011 and here we are at the start of 2017 and there will be no Solaris 12. Solaris engineering has only itself to blame - it hasn't delivered. I can't help but wonder if failing to deliver to a timeline may be part of what's happened here and Oracle management refuses to just keep giving money to an engineering group that can't deliver on time/budget.
There's also the question of the ISV market - if it wasn't worth it for ISVs to requalify their applications on Solaris 12 or they plain didn't see any customers asking if non-Oracle apps would run on Solaris 12 then that really kills Solaris. An example here might be IBM not willing to qualify DB2 on Solaris 12 and DB2 customers not interested in buying OracleDB. So either Oracle keeps Solaris 11 ticking over so that people can upgrade Sun hardware to run new Solaris 11 (and thus DB2) or it loses even more sales.
Solaris is dead. Start planning to move off it now if you haven't already.
Back in the day, the thing to do was run Oracle DB on Solaris.
If Solaris is effectively being EOL'd, what does this mean for Oracle DB?
Seriously?
As far as preferred platforms for things like Oracle database and other software goes, Oracle threw Solaris under the bus a little over 10 years ago, in favor of Linux. This unbent a wee bit after the Sun acquisition, but there was never any great effort to root out the anti-Solaris/pro-Linux sentiment within the company, especially in development and sales.
Other than some minor toeholds gained by valiant efforts of the SPARC/Solaris resistance, Oracle Cloud is all about Oracle Linux.
There is no denying that Oracle has invested significantly in SPARC and Solaris since the Sun acquisition and will continue to do so as highlighted in the latest version of the roadmap. You can easily dig up the SPARC roadmap that was first released in 2011, 6 years ago, and see that the SPARC and the Solaris updates have all been delivered, more or less on schedule and per the mentioned performance gains. Todays SPARCM7 is easily 11x more powerful than the first Oracle delivered T4 processor in 2011 and over the last 6 years, Oracle has delivered T4, T5, M5, M6, M7 and recently the S7 chip. 6 processors in 6 years! Not bad for a CPU product line that many are claiming is dead! Oracle over the years has hired 1,000's of engineers working on SPARC & Solaris and many of those folks are not even in the HW divisions as many of them are sprinkled across all of Oracles SW & Cloud platform R&D. Did you know Oracle spends over $5.5BN in R&D and its increasing at ~14% per quarter! Most of it is now focused on Cloud..
So these past few years, Oracle has clearly put a big stake in the ground and focused 100% on the Cloud. And this past year, IaaS has been the share of the focus as Oracle builds up its CLOUD datacenters around the world. Today, Oracle has 29 regions, where each region has several datacenters for disaster recovery and redundancy, and has plans to add 3 more short term and many more to come. Clearly, these datacenters will have a huge demand for hardware. And unlike most of its competitors, Oracle will be installing Oracle systems into its CLOUD, including SPARC systems, including Solaris. Not generic whiteboxes from the Taiwanese. So with the recent roadmap and reorganization changes, its clear that Oracle is aligning its resources to deliver more SPARC & Solaris in its cloud, both OracleCloud and at Cloud@Customer (Cloud on premise). So don't count SPARC or Solaris out just yet. Its been around 30+ years already and is clearly a survivor.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
