Comments on: Cloud Nine: Specification for a Cloud Computer. A Call to Action.

By: Fred Allsop

Fred Allsop — Thu, 19 Nov 2009 03:41:26 +0000

Great writing – enjoyed the read

By: Kimbro Staken

Kimbro Staken — Thu, 05 Jun 2008 02:39:46 +0000

Very interesting post, however I’m always concerned when I see talk of standardization so early in the development of a new technology. Obviously what you want is a very good thing, but trying to build it will severely constrain the amount of innovation that will be possible in the space. At this point we really don’t know what the right model is, Amazon does one thing, you guys do something else and Google is in an entirely different ball park.

Kimbro

By: Dave Rosenberg

Dave Rosenberg — Fri, 16 May 2008 20:03:38 +0000

@David—there already is an open source Tibco, its called Mule and we have hundreds of enterprises and multiple SaaS companies already using it as the integration layer for their siloed apps.

By: David Young

David Young — Thu, 15 May 2008 08:21:27 +0000

@Matt: oh well. About the links: we’re still running Textpattern here. We’d like to switch to WordPress…but we have lots of content. Regarding the software load balancers, I disagree. You don’t get layer 7, you don’t get fail over. Ghetto, Matt. For something that isn’t do much traffic, OK. But the cloud should allow for big scale.

By: Anti allergic

Anti allergic — Wed, 14 May 2008 21:04:31 +0000

Thank you for a long time looking for something like that.

By: Matt

Matt — Wed, 14 May 2008 18:58:21 +0000

I think software load balancers are more than enough for the scale you describe:

http://haproxy.1wt.eu/10g.html

(Links should auto-link.)

By: Benny H

Benny H — Wed, 14 May 2008 17:48:43 +0000

@Dan

Amazon allows you to pic the data center location of your node, which might help with redundancy.

See the following link for more information.

http://www.amazon.com/EC2-AWS-Service-Pricing/b?ie=UTF8&node=201590011

By: Dan

Dan — Tue, 13 May 2008 22:44:14 +0000

So far, your picture of the cloud doesn’t solve my simple problem: I want redundancy. If I buy 10 redundant virtual servers today, they could all be taken out by a single motherboard failure (if you allocate them on the same server).

There needs to be a way to ask “what’s shared physically?” in my logically shared-nothing stack. Hey, don’t you want your customers to buy a redundant server in each of your data centers? That needs to be part of your service (and eventually your API).

The statement about “loaning” infrastructure reminds me of the day that the NYSE found all of their expensive “redundant” connections (from different companies) cut by a single stroke of the backhoe. Turns out they were all traveling in the same fiber bundle. Real redundancy requires a physical view, not just a logical one.

By: David Young

David Young — Fri, 09 May 2008 05:04:02 +0000

@Steve: I concede your point. We have the API, it’s not public, yet. I gave Joyent a mulligan.

By: Steve

Steve — Fri, 09 May 2008 00:36:54 +0000

I count it as a score of 6/9, with the following missing from Joyent Accelerators:

- #2, API for cloning instances – #4, State Layer Interoperability – #6, Automatic Scale