I don’t spend that much time thinking cloud these days, although there are interesting machinations here at Google that might suck me in should I get bored with Android. But the topic isn’t going away just because I’m ignoring it.
Privacy and Cynicism · I remember being in San Francisco last year and there were ads on the taxicabs: “We care about the Private Cloud”. Say what? This whole “Private Cloud” notion is a conspiracy between CIOs who think they can do a better job securing data than professional shared-services operators (uh huh), and systems vendors who love the idea of selling enterprises way more hardware than they’ll ever need at one time, so they have the heavyweight infrastructure you need to support lightweight deployment.
It’ll Happen Anyhow · It might even be true that the business benefit of getting app deployments out of the clutches of Mordac the Preventer is enough to justify all the extra iron you’re going to need to offer Whatever as a Service.
It’ll happen anyhow because people will go off to AWS or Heroku or App Engine or wherever to get Whatever as a Service and, as with many other technologies, CIOs will be the last to discover that their business has already been bet.
I’m thinking mostly greenfield projects. It’s easy to see people building apps on idiosyncratic offerings like App Engine or whatever version of Rails or PHP your friendly local PaaS provider has on offer. But the notion of cloudifying production apps, which in my experience tend to have arcane tangles of highly version-specific dependencies, seems way tougher to me.
Where we going? · There’s a ton of outsourced infrastructure happening right now, and I haven’t seen much data on what’s popular and who’s winning. There’s a broad perception that Amazon Web Services are expensive in production, but people are using them anyhow. It’s a fun area to watch.
All this was brought to front-of-mind when I got a presentation of Stackato, being built right here by ActiveState, who’ve been around Vancouver forever. I liked the story, partly because it’s open-source-centric, which ActiveState has always been, and which infrastructure should by default be. And which also means that there might be a play for it outside the enterprise context.
Also, they have the radical idea that Rails and PHP and NodeJS and Django and Spring and even Erlang are all first-class citizens; love it!
Anyhow, I’m just happy to see that there are people out there with new ideas. Every year that passes, the notion of nonspecialist companies managing their own IT platforms seems increasingly insane at every level.
Comment feed for ongoing:
From: Sadique Ali (Jan 30 2012, at 10:37)
What is your take on the efforts put into this space by groups like Open Stack and Engine Yard?
Which approach- Open Stack's approach of do-anything-on-it cloud or Engine Yard's controlled environments would prevail in the long run?
From: John Cowan (Jan 30 2012, at 11:51)
I think that Private Cloud concept is being driven by the rather business-hostile SLAs of major cloud providers these days, which provide no indemnity if the service vanishes. As a simple example, the Amazon S3 SLA guarantees only uptime; Amazon is entirely free to throw away your data without warning, or to shut down the service altogether with minimal warning. At least your own cloud won't do those things on purpose, though they might by accident.
Migrating production apps should be straightforward on "anything-goes" IaaS services like EC2. I don't see these as being in competition with AppEngine vel sim.; they are a different mechanism for a different purpose.
From: Paul Hoffman (Jan 30 2012, at 14:10)
+1 to John Cowan's response, with a bit more. Given that no one can actually measure who offers better security, assuming that Some Remote Cloud Provider will be better or worse than a private cloud is just a guess. Given how unused CPU cycles get slurped up relatively quickly, over-buying for an internal cloud is probably not as bad of a blunder as over-trusting an outside vendor who can only lose business due to public shaming when they destroy your business.
From: len (Feb 01 2012, at 12:38)
What John and Paul said. Having managed a server center for a HUD application, security and SLAs were very important but security above all. They will bitch about downtime but a security breach meant losing the business.
One more nail in the coffin of the notion that openness is the only approach to systems. Evolution in hostile environments favors semi-permeability and private pools.
Otherwise, Chrome wouldn't exist.
So back to coding this XML editor using a framework from a vendor that did a good job of providing good XML components for just such occasions. Why? I need it to do what I need it to do. Sell it? No. Not enough people share those needs. They share XML.
Locale and scope.
Locale is everything when you consider the needs of the near neighbors to whom one provides services. Scope is everything when you have to decide how widely to share a value.