Slashdot Mirror
Microsoft, Amazon Oppose Cloud Computing Interoperability Plan
thefickler writes "Microsoft is opposing an industry plan, the Open Cloud Manifesto, to promote cloud computing interoperability. Officially, Microsoft says the plan is unnecessarily secretive and that cloud computing is still in an early stage of development,
but there are allegations that Microsoft feels threatened by the plan because it could boost Linux-based systems. The goal of the group behind the manifesto, the Cloud Computing Interoperability Forum (CCIF), is to minimize the barriers between different technologies used in cloud computing. And this is where the problem seems to lie, with the group stating that 'whenever possible the CCIF will emphasize the use of open, patent-free and/or vendor-neutral technical solutions.' Some speculate that Microsoft is actually worried that this will allow open source systems, such as Linux, to flourish, at the expense of Microsoft technology."
Amazon is also declining to support the plan, saying, "the best way to illustrate openness and customer flexibility is by what you actually provide and deliver for them." Reader smack.addict contributes a link to an O'Reilly piece asking what openness really means for cloud computing.
Microsoft... complains about something because it is too secretive? wasn't this in the book of revelations somewhere?
Obama is a twitter sock puppet
Openness implies lower barriers to entry. If they control the technology, they control the admission price. If you want to play on our 'cloud' then it's going to cost a CAL.
Sig this!
In the computer world, whenever there are a few entrenched players, the opposing companies often join together to try to counter their power, and they often do it in the name of interoperability and standards. So while in this case it looks like Amazon and Microsoft are the bad guys (let's be honest, Microsoft is always the bad guy), in reality it is just a matter of their competition trying to get a piece of the action. Who are the supporters of the CCIF?
IBM
SUN
CloudCamp
Zero Nines
and some others.
Similar to when Facebook started becoming the dominant social networking site, a few of the others got together to try to make a public API so it is easy for users to switch between sites. Typical corporate politics.
Qxe4
Maybe it really is too early to focus on standardization. It often freezes the standard quo and makes it harder to implement new stuff, c. f. the x86 and Windows requirements for backwards compatibility. I also don't really see where the problems are (others might have more experience there): EC2 uses standard Xen instances that should be somewhat portable. The only non-portable part is the meta-level configuration.
Fleur de Sel
Open Standards never work
So how did you manage to post that?
You don't have to maintain infrastructure to deal with your peak loads. You just have to keep enough to handle the baseline and than when you get hammered, you "turn on" more "computers" as you go. In theory, those "computers" could be located anywhere, so if you are mentioned on some UK news show and get hammered over there, you can "turn on" more of your "computers" to handle the load and turn them off when you are done.
In other words, basically, you have an infinite amount of computers which start almost instantly that you pay by the hour/minute for. Each of them boots off a standard image you control and all of the service providers have ways to script things like "hey, I've just been booted! lets tell the load balancer to add me to the pool!"
In yet other words, it is basically like a distributed virtual server. Take a single image and on-demand, load up as many virtual servers as you need.
First, the "cloud" doesn't run on anything. The "cloud" is basically a metaphor for an virtually infinite amount of servers you can fire up running your system image at once. It doesn't mean your instances are "floating" around a pool of servers--those images are running on real servers in some dudes rack and each running instance is indeed mapped to one server. If the physical server your instance is running on dies, oh well, you just fire up your image somewhere else. If you looked in the data center, you'd just see a bunch of regular servers running something like VMWare ESX (or whatever) and a bunch of fancy scripts to load and provision customer's images across the data center. You'd probably also see some serious SAN shit too.
All your instances typically connect to the same pool of shared, perminate storage. Each instance (at least on EC2) gets a couple hundred gigs of temporary disk space that goes away when you shut down that instance.
With Amazons EC2 (the only one I've played with), you can shove anything into your disk images has long as it is x64 or x86. "Anything" could be Windows Server, Linux, Sun, FreeBSD, whatever. You can download a lot of pre-build images from the community too--like "here is FreeBSD /w useful stuff already installed".
The trick right now is everybody has different ways to fire up said images. And once they are fired up, the API's your software must interact with are different. One guys way of provisioning an IP address or mounting a disk is different than another.
But this is to be expected. The whole industry is far to young to ask for standards.