4 thoughts on 4 clouds, from a guy who sees them all

3 Comments

Credit: Scalr

Sebastian Stadil has been managing cloud environments for years as the creator of the Scalr project and founder of the startup he built to commercialize it. It’s his business to know how the various public clouds work, and to understand how Scalr’s customer are using them. He came on the Structure Show podcast this week to share what he’s been seeing and how he thinks it bodes for the future of cloud computing.

It’s an interesting interview that touches on just about every cloud around, including Rackspace, IBM Softlayer and Google, and is heavy on speculation about how the OpenStack ecosystem will unfold. Here are four quick highlights, but anyone interested in hearing Stadil’s insights on usage trends and prediction should listen to the whole thing.

Download This Episode

Subscribe in iTunes

The Structure Show RSS Feed

On Amazon Web Services’ move into IT management tooling

“Having worked with Amazon for the past seven years or so, I’ve kind of gotten to the conclusion that Amazon’s just in the business of solving customer problems,” Stadil said. “And when it detects it’s IT that’s the problem, then they’re going to build tools to replace IT. When they detect that it’s something else, then they’re going to build that something else.”

On Oracle’s seemingly incomplete cloud

Stadil said he was experimenting and investigating Oracle’s cloud documentation and everything — storage, messaging, database — was looking pretty good. Until this happened: “Then I dug a little deeper and I found there’s actually nothing around compute. … Which basically means there’s no way I canget a virtual machine from Oracle.”

Structure 2011: George Gilbert – Principal, TechAlpha; Derek Collison – CTO, Chief Archictect, Cloud Division, VMware; Michael Crandell – CEO, RightScale; Issac Roth – PaaS Master, Red Hat; Sebastian Stadil – Co-Founder, SCALR

Stadil (far right) talks multi-cloud management a Structure 2011.

Windows Azure: Good tech, on an island

“Azure is going to be loved by developers the same way [Internet Explorer] is,” Stadil explained, in reference to the Microsoft’s tendency to use different terminology and standards than other software companies. “… I just see a lot of pain working with the APIs, and I think that’s what’s holding back Azure’s growth.”

OpenStack is still a zoo

“I would love for it to be boring, but right now it’s more like a rocket science project,” Stadil said in response to question about claims that OpenStack has gotten boring because it’s now so mature.

He continued, discussing how current open source technologies came to be because vendors tried commoditizing everything else in order to convince buyers to spend money on their particular part of the application stack:

“It’s kind of interesting to see how that might be playing out again, where all these vendors are pouring tons of dollars into OpenStack, now it seems like they’re doing the same thing to the Docker ecosystem. … The LAMP stack was largely a monolithic application stack and that became open source, it will be interesting to see how a new open source stack emerges specifically for distributed internet-scale applications. …

“I’m pretty sure that at the end of the day, we’re going to end up with an open stack that nobody has to pay for and all the vendors will have not made that much money on it.”

3 Comments

Jim Haughwout

Having used all of these since 2009, I would agree entirely. However, I could not sum each cloud up so elegantly in a phrase (e.g., OpenStack is still a zoo).

I would be nice to see some of your thoughts on Google Compute Engine. I have found it very useful for spinning up things like Cassandra nodes. I have also found the APIs easy to use and speed to be wonderful. Unfortunately as it is always in Beta, I have found it unappealing to enterprise and government customers.

sebastianstadil

Google has a clear second mover advantage, and they deliberately had long beta periods to allow them to get the APIs right. APIs are a contract, a promise between the provider and the consumer, and you cannot change APIs any more than you can break a promise or renege on a contract.

Amazon today is in a tough engineering position because they still support the APIs they released in 2006. If you don’t deprecate APIs, your mistakes are forever–and Amazon engineering today has to deal with the cumulative technical debt made by all of the mistakes market pioneers make.

Azure carries the legacy of their PaaS roots, and that is felt in their APIs. They know this of course and presumably are working to fix them. Ideally they would replacing them with a new non-backwards compatible set, clear of legacy constraints.

OpenStack faces a different challenge similar to that which politicians face: trying to please too many constituents and in the process creating a broken system. ISPs like AT&T want it to do NFV, Rackspace and CloudScaling/EMC want it to be like Amazon, etc.

Comments are closed.