Linux Kernels, Blog-o-Frenzy And Why You Should Care
Fall is in the air in New England and I am still sifting through the thousands of product announcements from Oracle OpenWorld 2010 (held in September). One of the questions I keep being asked by solution providers is “What does this new Oracle Linux Kernel mean to me?” Knowing how astute the VAR Guy’s readers are, I jumped at the chance to share my thoughts on this exciting announcement.
A colleague of mine wrote a great post on the press coverage surrounding how the introduction of Oracle’s Unbreakable Enterprise Kernel for Linux that he termed a “Blog-o-Frenzy.” It’s great to see the community so passionate about open source technology! The primary discussion points in the press seem to center around scary words like “proprietary” or “lock-in.” Since when does open-source equal lock-in? Isn’t the whole point that if the community does not like what one company, group or individual is doing, they are free to go in an entirely different direction?
In thinking about this topic, I remembered a recent blog post by Jeff Gould at Peerstone research titled Oracle to Red Hat: It’s Not Your Father’s Linux Market Anymore. I don’t agree with all of Jeff’s conclusions but completely agree with the assertion that “the non-proprietary nature of Linux distributions is a myth.” Going back to the premise of open-source being about choice and community development, Oracle Linux being branded proprietary because we are offering the community (in source and binary form) the latest kernel advances has me especially confused…is being open now considered proprietary?
How we got here
Oracle software, particularly Oracle databases, runs extremely well on servers with a lot of memory. As the price of memory has declined, our customers have taken advantage of this trend and deployed servers with huge memory capacities like the new Sun Fire X4800 Server with 1TB of RAM. Think about it: 1TB of RAM in an x86 server was unheard of when Oracle Linux 5 (or RHEL 5) originally shipped in 2007. During this same period, Intel and AMD made huge advances in x86 processor technology including 4, 6, 8 and up to 12 core processors later this year.
The Linux development community is quick to recognize those trends and optimize the Kernel to take advantage of those hardware advances. Unfortunately, most commercial Linux distributions have fallen behind the development trend. Generally speaking, kernel updates /advances have been reserved for major distribution updates, leaving users waiting to leverage all the capabilities of the latest hardware platforms. For example, Red Hat has been shipping the same RHEL 5, 2.6.18 kernel with security updates since March 2007, nowhere near the latest stable mainline kernel of 2.6.36.
Enter the Oracle Unbreakable Enterprise Kernel for Linux
In case you have not heard, Oracle is in the hardware business with the introduction of the Exadata Database Machine, Exalogic Elastic Cloud and the acquisition of Sun Microsystems. Our goal from the start has been to provide the best products, support and performance from application to disk through the optimization of hardware and software. Could we get maximum performance and reliability from Exadata, Exalogic or the latest Sun x86 hardware using Red Hat Enterprise Linux 5 or the Oracle Linux Red Hat compatibility kernel? In a word, no.
Starting with mainline kernel advances and adding data integrity features, optimized infiniband, a new version of the OCFS2 file system (and more) plus a LOT of testing, Oracle has introduced a new kernel for Red Hat Enterprise Linux 5 or Oracle Linux 5 called the Unbreakable Enterprise Kernel.
The performance impact of the new kernel is immediately apparent, up to a 75% performance increase on 4 socket, large-memory machines. Performance is only one part of the value proposition: The magic behind this new kernel is that all Oracle software is developed and tested exclusively using this new kernel on Oracle Linux in our development farms along with Oracle VM. Oracle is the only company with the resources to test this new kernel with enterprise databases, middleware and applications and ensure stability with 128 threads and a million IOPS. What does this mean to you? Oracle software is optimized to leverage the latest hardware, kernel advances and deliver the best performance and reliability on Oracle Linux.
Putting performance aside for a moment, one of the primary reasons ISV’s, System Integrators, Resellers, and Cloud partners should be promoting and using Oracle Linux is our integrated support and platform management. As customers or partners, do you prefer to be the coordinator of support relationships between applications, middleware, database, OS, hypervisor and hardware vendors? Would you prefer to make one phone call to Oracle to isolate and resolve problems? Oracle provides tuned, tested and supported solutions with integrated management to our partners and customers.
Putting the pieces together
Let’s go back to the question of open. If open means you can have the best performance, reliability and support then why wouldn’t Oracle Linux be your distribution of choice for your customers and in your own data center?
Want to learn more? A great place to start is the Linux Knowledge Zone on the Oracle PartnerNetwork. While you are there, get started on your Oracle Linux Specialization by enrolling in the free training including a module dedicated to the Unbreakable Enterprise Kernel.
Jeff Doolan is the director of virtualization and Linux channels for Worldwide Alliances & Channels at Oracle. Monthly guest blogs such as this are part of The VAR Guy’s annual sponsorship.
So the Var Guy is now sponsored by Oracle?
Ian: Yes, Oracle is among the site’s sponsors, as disclosed in the blog.
– Site sponsors are listed here: http://www.thevarguy.com/my-sponsors
– Additional site disclosures are listed here: http://www.thevarguy.com/disclosures
The VAR Guy is on record that he’s impressed with the OPN partner program to date but skeptical of Oracle’s open source moves and closely watching Oracle’s management of MySQL.
All that said: Any reaction to Jeff Doolan’s blog entry?
-TVG
Sounds like a closed system to me, if you want support. I can just hear Oracle support: “you did what? You replaced gcc? Your warranty is void!”
I can’t imagine ever using Oracle Linux unless running their DB, but that’s just the view of one Un*x veteran.
Wow. I think you need to put a banner on the initial links on your home page, or at the top of the article, saying it is a sponsored posting.
Reading the Novell one earlier in the week, I read half way through, and thought that VAR Guy was crazy in spending time writing about Novel, since no one truly cares much about what they sell any more. Then I thought the VAR Guy site had become an Oracle employee in reading this one, before jumping to the bottom and seeing the author.
I realize you need to eat, and realize it is this Oracle guys job to sell more of his stuff, and put the spin knob to 11 to try and get there. But it feels like those guys who used to be in EST (Erhardt Seminar Training – Sp?), if any others are old enough to remember that. Where they said “if you say a falsehood enough times, people will eventually believe it to be true”.
My favorite part is:
“Let’s go back to the question of open. If open means you can have the best performance, reliability and support then why wouldn’t Oracle Linux be your distribution of choice for your customers and in your own data center?”
and I flash back to the response I got from a CIO when Oracle announced the strategy around integrating Sun and one throat to choke logic. He said to me “That is all well and good, but when you are choking the throat and they are laughing at you and raising your prices, what good does that do”.
The only good news is that luckily the Internet era is moving past Oracle, and eventually their plans to drag everyone back into the caves will slow them some, and they will be a variant of Microsoft, or General Motors, or (pick the analogy of a lumbering cocky giant).
“If open means you can have the best performance, reliability and support then why wouldn’t Oracle Linux be your distribution of choice for your customers and in your own data center?”
Wat?
hahaha, never click links to http://thevarguy.com/ again!
You might wanna read a bit before you fall in love with OracleWorld FUD – a version numbers race is Linux 10 ago …
RHEL5 supports 2TB of RAM, so does RHEL4: http://www.redhat.com/rhel/compare/
The RHEL kernel is by no means years behind in enabling new hardware – even if the version number of the package might indicate it. Check hardware certifications or just Release Notes, e.g. http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5/html/5.5_Release_Notes
Watch the RHEL6 launch webcast – showing nicely how – while RHEL6 was worked – new hardware was enabled for RHEL5, performance significantly increased and power consumption reduced: http://www-waa-akam.thomson-webcast.net/us/dispatching/?event_id=09a0527bde2a763b87bdbe202ea8a5abamp;portal_id=af9b227bf07c733390c2738ee0330646
If you really wanna know, read the RHEL Advantage whitepaper: http://www.redhat.com/f/pdf/rhel/RHEL6_Advantage_WP.pdf
To put this up top: I work for Red Hat. Having said that, I’m speaking entirely in my personal capacity and not representing RH at all. Please don’t take this or report it as RH’s company line. It’s just me choking. 🙂
“The Linux development community is quick to recognize those trends and optimize the Kernel to take advantage of those hardware advances. Unfortunately, most commercial Linux distributions have fallen behind the development trend. Generally speaking, kernel updates /advances have been reserved for major distribution updates, leaving users waiting to leverage all the capabilities of the latest hardware platforms. For example, Red Hat has been shipping the same RHEL 5, 2.6.18 kernel with security updates since March 2007, nowhere near the latest stable mainline kernel of 2.6.36.”
This is, to put it mildly, claptrap. RH hasn’t ‘fallen behind the development trend’. Given that quite a lot of the code in those updates from 2.6.18 to 2.6.36 was written by Red Hat engineers, we’d have to be pretty schizophrenic to manage that trick.
RHEL 5 is a stable release. Therefore it doesn’t get version updates willy-nilly. This is not some kind of inability on our part to update it if we so chose, it’s just the sensible way to maintain a stable platform. With that said, it’s not correct to suggest that the current RHEL 5 kernel is just 2.6.18 plus security fixes; it’s considerably more complex than that. There have been significant feature backports from later kernel releases, often to meet specific customer needs, with extensive testing.
This post of course also conveniently forgets about RHEL 6, which was released yesterday; even if this column was written prior to that, I find it unlikely that someone in the author’s position would have been unaware the RHEL 6 release was pending. RHEL 6 comes with a much newer kernel, as you’d expect; I think it’s a 2.6.32 build, but I might be wrong (I work on Fedora, not RHEL).
If we want to talk smack, you could say to Oracle that it’s comparatively easy to just build a kernel and release it (hey, I can build a few hundred a day on my little desktop if you like!); the trick that I’d say RH’s track record proves it’s rather good at is long-term maintenance. That’s *why* RHEL 5’s kernel is so old, after all: we’ve been maintaining the platform, to the apparent satisfaction of our many customers, for several years. Oracle can’t provide that kind of track record with its kernel yet. Is Oracle still going to be supporting this kernel 2.6.33 build in five years? Will Oracle force its customers to update to newer kernel revisions to maintain support, even if these may be unnecessary and potentially disruptive changes for them? Or will Oracle have lost interest by then, discontinued its kernel builds, and gone back to using Red Hat’s?
“If open means you can have the best performance, reliability and support then why wouldn’t Oracle Linux be your distribution of choice for your customers and in your own data center?”
Open has never meant that… open means you let everyone else see what you’re doing. Just like it does in just about every other facet of life. Open is open.
While I don’t doubt Oracle’s additions and modifications to their “Unbreakable” Linux kernel and very robust and worth the effort; I don’t remember seeing any articles about Oracle “opening” any of that. And while I don’t follow the git log from the kernel, I think plenty of news sites would have picked up Oracle dumping that kind, and that much, code into the mainline.
@TVG
“Any reaction to Jeff Doolan’s blog entry?”
If you would point me to a blog entry, I would react to it. This is an advertisement, nothing more. All FUD, no facts. The only thing that even approaches fact is incorrect (RHEL 5 memory limits) If open means complete stack lock-in from hardware to web-ware, why not Oracle?
On record and disclaimered is one thing, but I’ve come to expect that blatantly sponsored drivel like this is marked as such so we can lower our expectations going in.
You know the risk here is that people will stop taking your blog serious.
Wow… what a complete waste of my time.
Just a quick note to say The VAR Guy is reading all the feedback and taking it all in.
-TVG
@Steven Oliver: everything Oracle has done to produce Unbreakable Enterprise Kernel is available in the open here: http://oss.oracle.com/git/?p=linux-2.6-unbreakable.git;a=summary It’s worth looking into what goes into mainline to see what Oracle has contributed and continues to contribute.
Oracle contributes a lot more to Linux than most people realize: http://www.oracle.com/us/technologies/linux/026042.htm
The source and binaries for Oracle Linux and the Unbreakable Kernel are available here: http://public-yum.oracle.com/ for the world to use/see (no subscription required)
Jeff
Oracle is (ab)normal thief – it sells things (Kernel, whole Linux distro – RHEL) which created others.
What a shame! Oracle == middlemen there is not invention at all!!
Also look at their poor DB – they presented 11g as a new – wtf? More than 70% is only what they called “improvement” – no, all this things were already prepared for 10gR2!
But of course, they need juicy customers. Sure, they will pay – they are stupid enough, they sure will pay…
No!!
Oracle, we won’t pay you bribe money!!! Stop to steal from others!!!
We already know Oracle amp; its openness – pfff…
As Openoffice.org? He bough it with Sun, close code, upset developers amp; lost many people.
Dear Oracle, money are good, but this is really bad time to cheat OpenSource community. We wish you fast doom…
Jeff,
It is very convenient to talk about a old release as just a kernel number but RHEL 5 kernel has a number of security fixes, bug fixes and even new drivers and other features routinely backported. It is even more convenient to ignore the brand new RHEL 6 release with a much newer kernel version and again one that would be maintained for 10 years. When Red Hat does the heavylifting upstream and in their product, you are essentially doing nothing more than a rebuild of a few srpms and removing the logos. This is hardly innovation. Oracle’s track record in gaining new customers with OEL is frankly dismal and a fluff piece here isn’t going to change that.
There should be a disclaimer on top of the post so people not interested in reading sponsored articles could avoid it. It was an unpleasant surprise getting to the bottom of the article to see that disclaimer. Thats all. Love the site.
I stand corrected then.
@Jeff Doolan
Contributions? You mean this one?
http://lwn.net/Articles/393144/
You ****tards tried to slip broken filesystem into Linux kernel, now you are calling contribution. Who knows what kind of backdors you might try to slip into linux next. All SCOracle idiots should be banned from LKML, and all patches from Redwood City should be rejected.
*NOT ACK*!!! I don’t think we need such “contributions”.
@OracleContributions?
I can understand frustration with business practices that don’t quite make sense from an outsiders perspective, but you’re being unrealistic; not to mention rude and vulgar.
Besides the article you quote ends with (and I copy and paste)
“chances are, this is not a case of a filesystem needing a fundamental redesign. Instead, all it needs is more extensive testing, some performance tuning, and, inevitably, some bug fixes.”
That sounds like every piece of software ever written to me.