Re: [clouds] Clouds and / vs. virtualization

[Paul Hoffman <[e-mail]>]

At 10:12 AM -0800 2/22/10, Tim Bray wrote:
>On Mon, Feb 22, 2010 at 9:56 AM, Paul Hoffman <[e-mail]> wrote:
>>
>> An earlier definition of "cloud" was a virtualized computer that was managed by an entity outside one's own enterprise. Later, "private clouds" were invented, which eliminated the "outside one's own enterprise" but also significantly changed the security model. Later still, "XaaY" (for many values of X and Y) latched onto the "cloud" buzzword, describing applications and services by themselves, not virtualized computers.
>
>I think there's reasonably good agreement on the values of X and Y.  Y
>is "S".  X is I for infrastructure, meaning roughly "like what Amazon
>does", P for Platform, meaning roughly "like what Google App Engine
>and Heroku do", and S for Software meaning roughly "like what
>Salesforce.com does".
>
>I'm not convinced that there's a value-add for the IETF here, but if
>there were, it'd be at the IaaS level.

I can buy IaaS; it is fairly close to the definition I proposed.

> > An IETF effort that tries to cover XaaY seems doomed to failure due to over-generality, while wasting a lot of participant effort. Instead, I think defining a cloud as "one or more virtualized computers where many components of each computer are managed separately" gives us a reasonable basis from which to work on how to provision such clouds and the networks over which the clouds run.
>
>The big value of IaaS is that it lets you fire up servers and
>configure them without involving Mordac the Preventer of IT, using
>only some Web forms and your credit card.  Amazon's API for doing this
>is kinda klunky, but tons of people are doing it anyhow so it's
>plausible that there really is a big market for this kind of stuff.

Fully agree.

>Thus there are lots of proposals for APIs that are better and don't
>belong to anyone, and a handful of efforts under way in various
>standards orgs out there; a few are enumerated at
>http://www.tbray.org/ongoing/When/200x/2009/04/28/Cloud-Standards and
>also check out VMware's vCloud.

Deciding where to start the protocol work should start after we decide what the protocol is supposed to be about. When we get there, there are many good starting points.

--Paul Hoffman, Director
--VPN Consortium