Slashdot Mirror
Does OpenStack Need a Linus Torvalds?
BButlerNWW writes in with a story that speculates about the need for a marquee name to head OpenStack. "OpenStack has been dubbed by some enthusiasts as the Linux of the cloud — an open source operating system for public or private clouds. But there's one stark difference between the two projects: OpenStack doesn't have a Linus Torvalds, the eccentric, outspoken, never-afraid-to-say-what-he-thinks leader of the Linux world. Torvalds personifies Linux in many ways. OpenStack doesn't have that one central figure right now. The question is: Does OpenStack need it? Some would argue yes. Torvalds, because of the weight he holds in the project, calls the shots about how Linux is run, what goes in, what stays out of the code, and he's not afraid to express his opinions. He provides not only internal guidance for the project, but also an exterior cheerleading role. Others would say OpenStack does not need a Torvalds of its own. The project is meant to be an open source meritocracy, where members are judged based on their code contributions to the project. OpenStack has been fighting an image that the project is just full of corporate interests, which is part of the reason Rackspace ceded official control of the project to the OpenStack Foundation recently."
It could be argued that one of the main reasons why Linux has utterly failed as an operating system for average people on average computers is Linus Torvalds. It has certainly been successful in other areas, but as a "just works" freeware replacement for Windows, it's been a bust.
Maybe it would be best if Open Stack stays relatively free of one person's influence, or one clique's interest, for that matter.
I've calculated my velocity with such exquisite precision that I have no idea where I am.
Comparing technical leadership to business leadership wow.
Linus essentially has one rule. Don't break it if it works. Even when Linus uses his bully pulpit to blast another project it is because they have broken things that worked.
Linus does not set a vision or a direction beyond code quality.
The comparison being made is like comparing an apples and a rubix cubes.
The summary makes a strong point: Linux is heavily dependent on Linus. Should we worry about this? What happens when Linus calls it quits, one way or another?
Mod me off-topic, if you must, but it's a question we'll need to face, at some point.
I started off writing this with; "While I have great respect and admiration for Linus I seriously doubt that having one unilateral "decision maker" is an advantage..."
After some thought it turned into; "Sure there is always a slow-down due to additional debate when there is more than one person at the helm however, when they reach a consensus and fail they will not be able to pinpoint the guilty party.
I guess what I'm saying is that if you have a wagon and that's pulled by 10 horses that's great, as long as they all pull in the same direction.
A 'singular oddity' is an event that cannot be explained and only happens when you are alone.
Linus specifically set up Linux development to not be dependent on him by creating git. People don't technically have to build from his own tree, but people do because they trust their experience with working with him.
You cannot just install a Torvalds into OpenStack. If there is no Torvalds of OpenStack, it's because no one is technically qualified or has the reputation for it.
This kind of reasoning is purely cargo cult management. You would think people have learnt to stop thinking in cargo cult ways by now.
Those who do not learn from commit history are doomed to regress it.
"OpenStack is a global collaboration of developers and cloud computing technologists producing the ubiquitous open source cloud computing platform for public and private clouds. The project aims to deliver solutions for all types of clouds by being simple to implement, massively scalable, and feature rich."
It seems to be mostly a Python-based automated system administration tool set for managing big machine farms. But the documentation is so buzzword-compliant it's hard to tell what actually works. The goal seems to be to have something like an open-source version of Amazon Web Services. Allocate real or virtual machine instances, load them up with executable images, hook them to some stored data sets, tell the network where the instances are and what they can talk to, and go. That's reasonable enough, and it would help if they'd just say that. And be clear about what actually works.
Do they need a self-opinionated little twat who writes three lines of code, bundles it with 300,000 lines of someone else's code and then names the whole lot after himself? No.
As formulated it's flamebait but seriously, you can't just pick a guy and nominate him to be the "Linus Torvalds" of your project and pretend it'll be the same. If Linus were to step down today, it doesn't matter who of the lieutenants who'd step up - they'd never have the same kind of authority to be the voice of Linux. It's the difference between being the founder like Jobs or Gates and your run-of-the-mill CEO. Even if your on top of the organization chart, you're not the benevolent dictator for life.
Live today, because you never know what tomorrow brings
I've looked over OpenStack and it looks great, with one exception: block storage (ie volumes.)
GlusterFS currently isn't recommended for VM storage by the GlusterFS people. They say "maybe" with the next release.
Sheepdog isn't recommended for production (and from what I read, provides abysmal performance - we're talking single-digit MB/sec.)
Lustre requires enormous setup+admin overhead.
DRBD isn't scalable beyond 2 nodes, really, and has serious issues with reliability and keeping in sync.
They've made a huge hullabaloo about Cinder - it's going to do my taxes, slice bread, and surpass Christ - but information as to what the hell it actually is or how it'll do it, beyond marketing-speak, is difficult to find. If you dig around, you find that it's a layer on top of other network block devices.
Far as I can tell, the only free (in either sense) backend they support is Sheepdog, which, as I said before, isn't considered anywhere near production ready.
It also appears that 'Highly available', 'fault-Tolerant', and so on- is coming from the underlying storage, not Cinder itself.
So, where's the beef? You can't have an "open" visualization system if you then require a netapp, IBM, or nexenta backend (sidenote: has anyone SEEN nexenta pricing? Holy christ on a stick!)
Please help metamoderate.
Awww, did someone's widdle kernel patch get rejected...?
I have no idea. With something like Proxmox I download it, install it and start running KVM and OpenVZ machines. Easy. With OpenStack, I go to their web site and I find nothing but a bunch of marketing crap. Cynical me just looks around there and thinks that some companies have got together to make something look open and look as if there might be some open source code and downloads 'somewhere', but there aren't. This is all to try and protect their expensive 'real' products that they know are probably under threat from a truly open source competitor but they just want to muddy the waters.
I think Joel Spolsky or someone once called it 'fire and motion'.
Linus made a project to make a UNIX like operating system. He found like minded individuals that wanted to help on USENET. I still look fondly back at the old days of trying to get the build to work. Later own I was able to have CDs mailed to me from people like CheapBytes that made the work so much easier (I don't missed the days of dialup internet). Eventually RedHat came around and I was hooked. Through it all Linus kept the project on course.
It takes a lot of discipline to take an idea from a post on USENET in 1991 to what Linux is today. His discipline and stewardship is worth way more than any code that he contributed to the cause.
Okay I used up my "stick up for Linus" allowance for the year.
These comments are my own and do not necessarily reflect the views or opinions of my employer or colleagues...
The KVM bits do seem to be most tested. The Xen stuff works, people use it, but I do question if it is as polished.
CloudStack supports XenServer very well, but it also suffers from all of XenServer's architectural faults and many of its own as well.
(Xen itself is well architected, in my opinion, but the closed XenServer introduces a few oddball design patterns that made sense in a small rack deployment that aren't good for scale out patterns)
Actually Linux did not start out as a "serious system" it started out as a hobby. I took the liberty of googling for an old post of Linus to show the understatement of the last millennium:
These comments are my own and do not necessarily reflect the views or opinions of my employer or colleagues...
I've been running vanilla Debian stable for about a year and a half after finally giving up on Ubuntu. Before Ubuntu, I was running Gentoo. Before that, Mandrake before it became Mandriva. Before that, Red Hat. Suffice it to say, I've got some experience with a few varieties of Linux that date back about 15 years.
I haven't had to drop to the command line for anything for a long, LONG time. Not a single admin task, application, or quick function. Not once. Virtually every capability has been available as a solid, dependable GUI app for close to 5 years. The last couple of laggards were probably audio and video driver management. Even those had decent GUI alternatives about 3 years ago. Heck, Debian's graphical installer worked just fine for a newbie that I introduced to it.
Like you do on Windows, I still go to the command line for some stuff. It can be faster to do so in some circumstances, for one. For another, I happen to like some of the console based apps a lot more than their GUI counterparts, but that's at least partly due to my familiarity with the CLI version than it is the GUI interface itself. Also, automating stuff is a heck of a lot easier and more flexible with a shell or Python script than it is through a GUI.
The point is, if you still think you need to drop to the command line to do anything on a Linux box, you are seriously out of touch with the current state of affairs. (Of course, if you've only tried Ubuntu I don't blame you for being confused. It's a lousy distro for a lot of reasons.)
You know what, in the interests of catching up with things (because you're absolutely right - I am out of touch and most of my experience comes from Ubuntu), I'm downloading Debian 6.0.6 right now. I'll throw it in a VM* and see how it plays out.
Amusingly your linux history isn't too different from my own, aside from Red Hat and Gentoo. Every now and then I give linux another shot and usually stumble upon some small, deal-breaking issue (Missing wireless driver, or missing SATA driver - dumb stuff like that). However, it has been a while so I'll give it another shot.
* I'm grossly aware of the limitations of a VM, do not worry.
+1 IDisagreeSoHeMustBeATrollOrAnAstroturferOrAShill