Slashdot Mirror
The Real Truth About Oracle's 'New' Kernel
An anonymous reader writes "Yesterday at OpenWorld, Oracle announced a 'new' Enterprise kernel for its so-called Unbreakable Linux. What's the real truth? The company is simply sticking a 2.6.32-based kernel on top of its re-branded Red Hat Enterprise Linux clone and trying to spin it as a new and innovative development."
This Barbie has a new hat!
1. 2.6.32 Kernel
2. re-branded Red Hat
3. ???
4. Profit!
Don't a lot of Oracle products just have Oracle stickers on top of some other product line they acquired anyway.
But...but...but...Theirs is unbreakable!
I thought this was perfectly normal in many industries. Remember the SNES CD addon that became the PS1? And how 3G and 4G mean whatever the company wants at the time?
It's Malabu Stacy with a new hat!
This could be glossing over quite a bit of useful work for Oracles customers. "Fine tuning" could be anything from tweaking some compiler settings to actually patching things in the kernel. Its hardly a trivial task given the size and complexity that most Oracle customers bring.
If the idea was to cause panic or start a conspiracy theory, it failed miserably. Nothing to see. Oracle is simply making a new kernel available which is newer and has more enhancements. Instead of waiting for RH, they are taking control of that piece of the distribution (if customers want it). Oracle should do the same with the rest of the OS and try to innovate there, instead of simply distributing pristine RHEL with their logos. But then, they already have Solaris which is much more suited for the markets they are aiming at (high-end enterprise servers), so why waste the time ?
none
What did you expect, that Oracle will have coded their own kernel from scratch? Every distro uses a version of the same Linux kernel. TFA is a troll.
Caveat Utilitor
We see what you did there, oracle...
Good people go to bed earlier.
They made direct comparisons to RedHat kernels claiming performance, security and stability enhancements? If it is the same, then those claims cannot possibly be true. This is confusing... and troubling.
Just one example of why this is good - iotop.
I've been watching the RHEL bug for adding iotop since at least RHEL 5.3. It keeps getting bumped, now RHEL 5.7 IIRC.
It would require a bunch of backporting work from the kernel beyond 2.6.18. But once sysadmins get used to knowing which disks are busy they really get used to that. And doubly so for optimizing database servers.
Redhat's strategy gains them certainty and loses them opportunity. That's certainly a niche that's done well for them, but there are also users with other needs. Oracle's strategy will be very popular with some of them. When Redhat brings RHEL6 to market there will be lots of required subsystem changes to get the new kernel. Some people will just want the new kernel and not want to change all their underlying dependencies, and Oracle is meeting that need. Eventually Fedora will adopt a rolling-release model and RHEL will track that (probably with more QA) but it's a hard problem and not well-solved yet.
It's great that we have such a vibrant market that there's room for so many approaches.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
To innovate means to make something new happen. It doesn't have to be radically new, just something that wasn't available before. In the real world, most innovations are pretty humble, but humble doesn't imply not useful.
Do you ever look at Crapware 7.0 and think they just added some 3D arrows for absolutely no reason? Now look at TFA and the reactions here, this is *precisely* why the marketers demand idiotic features.
If you've actually set up Oracle on a system, you quickly realize that a. it's hugely complicated but b. it's a solved problem so c. why am I going through all this pain when Oracle has done this already? Of course, they have, calling it OEL just makes it easy to explain to the boss.
And for anyone trying to maintain an Oracle system, this is a big deal. It is not an understatement that for the typical business, their Oracle database more or less *is* the business. You want something that's going to work, with no nonsense, and you want to keep it up to date.
Oracle Linux is Unbreakable and better than Linux.
BUT Linux is bad mojo.... if you want a real OS and not a toy, use Oracle's Solaris.
Somehow they failed to add that last bit. Mixed messages from a VERY mixed up company.
People may want to check the LWN discussion on the topic, which includes comments from Chris Mason and others concerning their improvements over vanilla 2.6.32:
http://lwn.net/Articles/406242/
I think this just fits in with the corporate IT empire building that has been going on lately. HP and Dell just finished a bidding war for 3PAR and now Oracle is making noise about "their" updated kernel. Just another enterprise IT infrastructure trying to stake its claim.
Whenever a company starts calling their product unbreakable or indestructible or unhackable or whatever, I start thinking Titanic.
My postings are informational and does not constitute legal advice. Act on it at your risk.
They specifically state during the keynote that the announcement is an update to the new kernel with some enhancements for their own use.
This post is a moronic display of the op's inability to read.
And here I thought they'd only just adopted Hurd.
Editor Emeritus and Senior Writer, TeleRead.org
this kernel is not the same as RedHat's, there are improvements geared toward Oracle's DBMS
oh please, oracle customer complexities are a result from the oracle usage and not the motivation for it.
Oh wow, what a revelation. Using a complex software causes usage complexity. Here, have a banana as a price.
Yeah, usage of Oracle causes usage complexity. Does that mean that fine tuning a Linux distro to ease the pain of configuring a box suitable for Oracle products is something trivial, or non important, or what? What was exactly the point?
It doesn't even have to be for running Oracle database-related problems. When you run a EE container, be it JBoss or WebLogic (now a Oracle product) on a HP-UX, Linux or Solaris box that sits between a HTTP server and a database server, you are still bound to tune it for efficient performance according to the specifics of the system. I cannot think of anyone simply dropping a box with software on it on production without the necessary configuration.
That configuration is repetitive, tedious and specific for any non-trivial product for non-trivial usage. It is hardly an Oracle side effect. Typically sysadmins have to automate those configuration changes (or keep a golden ghost pre-configured image.) No matter what, that is still a burden. Better yet to have a vendor backing a set of configuration items already packaged into a turnkey solution.
oracle is one of those business providing useless solution so they can charge you twice for the consultancy.
Just because you don't like it and like to apply partisan ideologies to engineering, that does not mean that what they do is useless. It might be useless to you, might be useless in some (actually many) business contexts. But that does not mean anything on the general case where having an Oracle solution (not just an oracle database) is a useless solution.
Engineering != rhetorical bile.
We bought the support from them. No penguin.
So, let me get this straight. Oracle is "bad" because they announced that their distribution had a "modern" kernel, but it's "only" 2.6.32 with custom patches, not 2.6.35 which is totally almost 2 months old now so there's no excuse for it not to be in there!!!! And, Oracle is a jerk who just takes and takes without contributing back, because they are "only barely" in the top 20 contributors to the kernel (and the kernel is only one small part of Linux so basically they don't contribute at all). What a troll! At least the article is up-front about being written by a Novell employee. (Wait no it's not, it sort of slips that into the middle).
And Mr. Sour Grapes Novell employee is just pleased as punch over pointing out the "dirty secret" Oracle tried to hide, by publicly announcing that Oracle Linux would be running the 2.6.32 kernel, with custom patches to improve performance on certain hardware, and for Oracle software. How sneaky of them, you could never tell by reading that, that it's actually the 2.6.32 kernel (WHICH IS SO OLD HOW DARE THEY CALL IT MODERN).
ASCII stupid question, get a stupid ANSI
Wouldn't they be sticking the new Linux kernel *under* their distribution?
"Ayn Rand is a bloody socialist compared to me." - Robert A. Heinlein
This could be glossing over quite a bit of useful work for Oracles customers.
You are glossing over the point of the article.
1. Redhat writes lots of great Linux stuff that make the kernel better (11.6% of the kernel).
2. Oracle passes it off as their own. (They only contribute 1.3%, less that 1/10 that of Red Hat).
I'm not entirely sure why this story is even here (full disclosure: I work for Oracle). Oracle Enterprise Linux has always (admittedly, braggingly) been a RH clone with support provided by Oracle and various additional packages bundled in ... not news. Initially, I assumed Oracle was targeting RedHat for a takeover (they still may be) and this was an attempt to financially weaken the company before the hordes descended, but it now looks like maybe there were other motivations. In any case, the new kernel is different than the one that RH ships (you can run either the RH kernel or the new one shipped by Oracle) and it's been optimized for the sort of load that Oracle thinks their servers will encounter. Great! Hope it works! Why is this news? How did plainly available information get spun into some sort of conspiracy?
When will Oracle be paying the Microsoft license?
And the fact that people actually grade the post I'm replying to as insightful shows the actual degree of industrial exposure sported by many of these /. fanboys. I mean really, can't these fools think in engineering terms without choking into whatever flavor of pseudo-liberating koolaid currently en vogue?
Yesterday's slashticle already specified that they put a 2.6.32 kernel in. Nothing to see here, move along.
What a depressingly stupid machine.
Anecdotal evidence, but where I work there were some people using pro*fortran to access Oracle databases from Fortran. pro*fortran was dropped between Oracle 8 and 8.1
It took six months of digging for the Oracle support people to finally tell us they had dropped pro*fortran from their product. Everyone kept saying "sure, we support Fortran, but that's not my specialty, let me get an expert for you"
When the technical support people don't know their own product, what worth is it paying for that it?
Sure, you can use OEL for anything you might want - but, the folks using this are probably folks using Oracle for the OS, applications, and possibly even hardware. What this means the Oracle *applications* are going to have better support and tuning.
The big news from Oracle is that it's offering a "modern" Linux kernel that's supposed to offer better performance and support for newer hardware (like solid state disks), and is optimized for Oracle hardware and software.
In practice, it works out something like this. Lets say you call up with some sort of goofy DB or Weblogic issue. Support *has* your exact environment. The application developers may have also used that same environment for development, making this the 'native' build rather than some other platform that the codebase was ported to. It also sets the bar on what you can do with some of the newer kernel features. Sure, you could custom tweak your own kernel to get some goofy bit of hardware to work, but if it breaks the app and you have to call support... Think of it as more of a least common denominator for the Oracle dev folks.
RHEL, OEL, and CentOS are all the same bloody codebase. Thank $DEITY. Pick your support contract vehicle on the commercial side. The fact that commercial applications run very nicely on that cut of Linux is one of the reasons Red Hat has the following it does. (I've got Red Hat in my dead pool for companies to be acquired - I'm surprised something this strategic remained third party this long.)
+++ UGUCAUCGUAUUUCU
Why isn't there a greed tag. While Oracle and greed go hand in hand?
Someone has to take up the vacuum left by SCO, I suppose
There are things Oracle could use if they really wanted "unbreakable". There are some very tough microkernels available. LynxOS is certified to DO-178B leval A for safety-critical software, yet it can run Linux ABI binaries.
LynxOS drives quite a number of systems with serious firepower. The Navy Shipboard Self-Defense System, the "Multiple Missile Kill Vehicle", stuff like that. On the civilian side, LinxOS powers the Airbus navigation system.
There's a performance penalty over Linux, and LynxOS is not free. But if it really has to work, there are options.
The article is missing the point. The key pitch here is that you need to be running Oracle Enterprise Linux 5 Update 6 or newer to work on the latest generation of Sun x86 hardware. It's a big deal inside of Oracle because Oracle wants to be running on Oracle hardware, but is about 80% Dell stuff on the x86 side right now in the Oracle data centers that weren't Sun acquisitions. There's a substantial hardware refresh effort inside the company right now, temporarily making Oracle one of Oracle's biggest hardware customers.
But this is part of a pitch to existing customers: run our OS and you have full hardware support TODAY. Run Redhat and you'll have it when they release Redhat 6 or if they decide to backport new hardware features to their kernel in a few months or years. The announcement is a statement that Oracle -- for the first time -- is taking the lead in releasing a newer kernel ahead of Redhat, rather than waiting for the Redhat release first before releasing the slightly-tweaked-for-Oracle version in Oracle Enterprise Linux. It's driven by hardware needs, and for at least several months that will be a selling point to customers wanting the latest and greatest: use OEL4u5 or Solaris, not Redhat, or else you won't leverage the new hardware features effectively (if at all).
I actually think the compatibility issue may just boil down to the SAS driver update to work with Sun's latest chipset. But it's a bit of a show-stopper if you're not running OEL5u6, since you can't even install the operating system without the SAS driver update.
Matthew P. Barnson
I learn what I think when I read what I write
Well, Linux is not new and innovative either, it is a rehash of Unix. It is amazing what you open source zealots believe.
It's called Fedora
Yes Francis, the world has gone crazy.
I really wish there was some way that Apache could have bought Java instead of Oracle. I know they don't have the money, but no one else would have taken as good care of Java. (And yes, this is a bit off-topic but all this is endemic of Oracle's attitude toward open source)
If Linux Had Modern Corporate Marketing maybe it would be in wider use. You make fun of it, but the masses of people respond to this kind of marketing. And by 'respond to it', I mean they buy more of the product that the bullshit marketing is, well, marketing. Microsoft does well because of it marketing. Apple does well because of its 'a light will shine out of your ass if you use Apple' marketing. Make a product look cool to use and people will come. The masses are like small children who watch a commercial for some hard plastic doll that is made to look cool and how you will be cool if you have one, and when you get it home it is just a piece of hard plastic. But they will learn to make it cool because no one wants to admit they just bought a hard plastic doll that doesn't do anything. If marketing were done to change the image of Linux from being the OS of glasses with tape between the eyes wearing pocket protector Asperger geeks, to the OS of rock star cool surfer dude hollywood shaker, more people would use it. The more people use it, the more money will go into it. So make fun of Oracle if you want, but this might be the kind of approach needed. After all OS2 in its day was far better than Windows 3.11 and Windows 95, and we all see what marketing it to the 1337 techies and "IT professionals" got it. If you never heard of OS2 then you *really* get my point. And Windows 95 was marketed using music from the Rolling Stones, and we were all told we would fart rose petals if we bought it... and it was a hit from day one; even if IBM/OS2 had a ton of time to capitalize on MS's major delay on the 95 release date... good/sleazy marketing sells.
-- I ignore anonymous replies to my comments and postings.
Hmm... 2.6.32. Did they fix this bug:
https://bugzilla.kernel.org/show_bug.cgi?id=16991
That is new and interesting for an "Enterprise Linux".
RedHat and CentOS still use 2.6.18. That's 4 years old - ancient, relatively speaking. We're talking CFQ with ext3 filesystems, which is a complete nightmare in terms of performance.
There have been quite a few improvements since. So yes, using a modern kernel on RedHat derived stuff is indeed "revolutionary".
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
each new solaris release and even many of the kernel and interface patches add support for several Intel controllers at a time. So it's not like this isn't addressed and is totally ignored. These are RFE enhancements under the dev system and they get incorporated either b/c the support is needed for a feature OR there has been a sufficient number of customers added to the RFE record. That's how it's done. It's not a conspiracy or anything like that. Is your company listed on the RFE report?
Oracle has a long history of running RHEL with FC kernels.
The focus should be on open source software, not some companies specific magical distro "stack". Everyone needs to push for modular and universally-installable software across all distros, and GNU/Linux needs to be its *own* distro. Then you can turn to that company claiming they are somehow magically superior and ask them what software they are using or configuration changes they have made that makes them supposedly so secure.
Promote true freedom - support standards and interoperability.
Oracle is doing what it's doing best: being Oracle.
Seriously, I have only seen bad decisions come from this business lately. How they managed to make this much money to get there I'll never understand.
I am not devoid of humor.
And yet I've seen plenty of ASP.NET books. .net software stack. I've never seen a commercial for Active Directory either. Your arrogant attitude doesn't improve Linux in any way.
Don't underestimate how much MS can attract those would-be developers with the
and you won't have any problems.
Oracle: No problem, here's the source.
Community: You need to contribute more to the kernel.
Oracle: I thought we only need to follow the license?
Community: That only applies to companies we like.
Don't fret, word is native ZFS is coming to Linux next month.
Oracle basically said to service their high end Linux customers the kernel in the current RHEL offering is out of date, so they decided to support 2 kernels for their customers... one that tracks the RHEL kernel (Unbreakable Linux) and one that tracks the mainline Kernel (Enterprise Linux). Enterprise Linux includes all the high performance features Oracle needs that even the mainline Linux community is too slow to incorporate. That's all Oracle is doing. Sounds like a perfectly reasonable approach to me.
Almost everything Red Hat ships in Enterprise Linux is not from Red Hat.
Excellent point... there are 100 other Linux distributions, and I haven't seen any of them approached with as much /. FUD as I've seen in this thread.