Drivel dripping with sarcasm in a one post tumblr blog. I really hope this crap doesn't get any more upvotes.

>> The competition and choice is essential to the open web. Yet OpenStack wants to be the only game in town?

Yeah, because CloudStack and Eucalyptus are just going to disappear overnight. What do you expect from a presentation given at the OpenStack Summit? "Here's Why CloudStack is Great?" Ya, competition is good so if CloudStack disappears tomorrow you can create your own competitor, what's stopping you?

>> No, absolutely nothing has been commoditized.

Ok.

>> So if I pick up OpenStack or OtherStack they should be basically interchangeable.

They aren't? Given your example of a HTTP server I'm sure I can just slap my httpd.conf onto a ngix server and off I go, right?

>> I’m sorry, but I missed that term in my Operating Systems 101 class. Why? Because there is no definition.

You must have went to a school where you memorized terms and didn't learn to think for yourself. The idea is that OpenStack sits on top of hardware, the same way an OS sit on top of hardware. If you want to get all grammar nazi about it you won't score any points with anyone.

>> So somewhere, at some point, OpenStack completely lost focus and now solely focuses on promoting itself. As though the original problem was that we didn’t have a thing called OpenStack and we have now filled that gap by creating OpenStack.

What does this even mean?

>> So can we get serious now about commoditizing the cloud?

Yeah dude, Amazon, Red Hat, HP, and IBM haven't already done this or anything.

>> I have some thoughts there…

Sure, I'll be back for your 2nd post but I hope it has some actual substance.

Totally agree.

I think the problem is that the definition of what a "Cloud Operating system" is actually is quite loose. Mine would include compute, storage, and networking.

And truly, by this definition, OpenStack did achieve something and filled a gap, as it integrated all these components (and so did CloudStack / Eucalyptus).

Wikipedia opens up to describe an operating system as "a collection of software that manages computer hardware resources and provides common services for computer programs".

Since the majority of the programs we build today run on servers, not desktops, we needed new APIs to compute / storage / network. A new class of OS provides these, which we call IaaS or Cloud Computing. Some of these new OSes also provide the aforementioned 'common services' such as datastores or event loops.

I think the excitement lies around having an open source OS for programming your infrastructure to, instead of a (nevertheless excellent) proprietary one.

I share your criticisms. This article isn't worth the bits it's stored on.

I think the main goal is not to make cloud software the commodity, but rather to make IaaS a commodity, so that there will be multiple vendors of IaaS that are largely interchangeable.

There are today a number of public clouds that run OpenStack, of which HP Cloud and Rackspace Cloud are the two most visible. Because they're all running OpenStack, they are (hopefully) going to be interchangeable commodities. There's still some way to go, but we're much further along than we would be without OpenStack.

I believe also that having multiple implementations is not discouraged (but is not the primary goal). The APIs are distinct from the implementation and you are welcome to re-implement them (the docs make a distinction between e.g. the Identity API and the Keystone Identity implementation). I do think it would be interesting to experiment with alternative implementations. Many people, for example, replace the Swift storage component with Ceph, as both support the Openstack Object Storage API.

I don't think commoditizing the cloud, in the sense of attempting to make all cloud providers ferature-for-feature compatible, is the right goal either. Sure, ideally we'd like if it possible to easily migrate our entire stack from one cloud host to another with minimal disruption, but frankly it doesn't happen that often, so that's too narrow a use case for OpenStack to try to solve.

Instead OpenStack's intent is you have a bunch of hardware in a datacenter, then you use this software to make your own cloud on that hardware. That in turn opens up the market for more cloud providers. Obviously Amazon offers the most mature feature set in that space so trying to offer similar features as Amazon is a worthy goal.

I thought the NIST definition for cloud computing was fairly succinct. The five essential characteristics are:

  1. On-demand self service
  2. Broad network access
  3. Resource pooling
  4. Rapid elasticity
  5. Measured service

I'm fairly certain Openstack as well as Eucalyptus and Cloudstack all meet that definition.

Yup. And you can add Open Nebula to the list.

BTW, given that NIST uses Eucalyptus, you could argue that Eucalyptus meets not just the formal definition of cloud computing, but also the practical one.

It's an interesting start but seems unfinished. I'd like it to go into more detail/support of the arguments.

A couple of problems with this writeup.

First:

Actually, commoditization is about when an atomic capability becomes tradable/sellable:

GOOG says "A raw material or primary agricultural product that can be bought and sold, such as copper or coffee." (http://www.google.com/search?q=define+commodity)

Coffee, copper, coal, and oil, all are differentiated. There is light sweet crude, heavy crude, and others. Copper from different mines has different properties and quality. Coffee is highly variable. Yet all are commodities.

The base unit doesn't have to be exactly the same, it just has to be measurable in terms of value so that you can compare one against the other.

This assertion has been corroborated in recent debates I had with an ex-commodities broker who was very clear that the commodities he bought and sold were quite frequently differentiated and variable.

For those who care, that was Dr. James Mitchell, CEO of CloudOptions, who was previously Global Head of Commodities for Morgan Stanley. Blog postings here: http://www.cloudoptions.com/blog/

Second:

This article's argument is essentially that if OpenStack is the winner, then everyone else loses. This is not a zero sum game, so there can be multiple winners, but just as with Linux, it's likely that there will be a very small number of winners and those will mostly dominate and then there will be a very long tail.

Linux success (as you can see in my original slide deck) meant that UNIX and other early x86 UNIX derivatives (e.g. SCO UNIX, 386BSD, FreeBSD, OpenBSD, NetBSD) essentially "lost". But Linux "winning" meant that we have a relatively standardized server operating system running globally, which actually increases competitiveness, allows for people to focus their learning and knowledge, and makes everyone's life better. The conclusion that is leapt to isn't really supported by any evidence. It's simply asserted.

I say this as a long time BSD guy. I wasn't happy that Linux won, but I am happy that we have a relatively standard server operating system.

Finally:

To imagine that this will be any different with the cloud operating system wars is nonsensical. Clearly there will be one or two major winners and OpenStack is positioned to be the primary winner at the moment.

Also read this:

http://www.michaelnygard.com/blog/2009/02/fast_iteration_ver...

Fast iteration wins. OpenStack is far ahead in this regard and only picking up speed.

Cant' you use open stack as part of a virtual private cloud that includes AWS?

It won, partly, because is has nothing to do with Java App Servers.)