On Jun 12, 2011, at 2:48 PM, [email protected] wrote:
> I'm on a new mail client and accidentally responded to just Derek.
Heh, then ignore my "direct to you" reply. :-)
> while a good pxe setup is great, and really, the only reasonable way
> to manage a fleet once it gets beyond a certain size, I think it's silly
> to call it "the cloud" because that sort of thing has been the way it's
> done at larger places for most of my career.
Well, we can have that whole discussion about the word "DevOps", too, no? It's
just a new name for the same flavor.
> So yeah, instead of 'internal cloud' I say "automatic provisioning system"
> I mean, it's a good thing to have, essential once your system
> gets beyond a certain size, but saying that this has anything to do with
> the decision to virtualize, I think, is a mistake. If you do virtualize,
> you should put effort into making the virtuals provisionable through the
> same system as physical servers. Heck, cobbler does this out of
> box.
>
> I think one great disservice all this 'cloud' talk does is relating
> auto-provisioning to virtualization. Both technologies are great when
> you need them, but you can have auto-provisioning without virtualization
> and vis-a-vis. Personally, I think that automatic provisioning is
> useful in more situations than virtualization is, but really you
> should look at both technologies on their own; You are doing yourself
> a disservice if you virtualize because you want automatic provisioning
> capabilities.
I both agree and disagree, in part.
I think that a true "cloud" system requires one of two things:
- a PLETHORA of different hardware configurations hiding inside the
cloud implementation, or
- Virtualization
Because at the end of the day, cloud computing, to me, is about the ability to
say "I need a new server, with 2 CPUs, 8GB of RAM, and 40GB of disk, running
$SOME_APP", and the cloud just "puts" that application somewhere inside it. You
neither know, nor care, what actual hardware it's running on, where it's got
the disk storage allocated, etc.
In a non-virtualized environment, that requires you to have a whole mess of
different configurations, depending on what you're going to get requests for OR
it requires you to have a whole passel of application servers that are getting
assigned hardware wayyyy over their use-case because you're spec'ing the
hardware to the maximum use-case.
In a virtualized environment, you can literally just let "the cloud" put your
2CPU/8GB/40GB config on the same hardware as a 6CPU/2GB/200GB cloud-based
appserver, and not care. So long as the cloud is managing those resources,
shifting servers around to account for load, etc., it's completely irrelevant
to you. And that, TO ME, is the hallmark of cloud computing.
Implementation issues - like how do you do the osload, is it via PXE, or
cloning, or whatever, is largely an irrelevant issue in that respect (For me at
least).
D
_______________________________________________
Discuss mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
http://lopsa.org/
