Cloud Computing: Follow-up

I received a couple of follow-up email messages on the last post and want to share the author's additional insights. The responses are listed below as received.

From Justin Dugger: I appreciate concept, but these examples aren’t really highlighting the “compute” in cloud computing.

A better example might be Replacing a Pixar renderfarm with Amazon’s EC2 service (http://aws.amazon.com/ec2/). EC2 allows you to basically rent time on Amazon’s massive cloud, and do so on demand. For example, you might when a computer modeling class wants to render a scene, rather than buy a few dozen computers that sit idle, you might buy time on EC2 and use a few thousand computers for a brief moment instead. When it’s done, those computers become part of the “cloud” again.

EC2 is marketed towards websites, where “flashcrowds” (http://en.wikipedia.org/wiki/Flash_crowd) can overwhelm organizations not expecting a sudden spike in popularity. By using EC2, you can snap up more computers and bandwidth instantaneously to serve that traffic, and relinquish it when it dies down.

Cloud computing or not, your point about data loss and security on the web is well worth highlighting. There’s an New Yorker cartoon about the internet: “On the internet nobody knows you’re a dog” (http://en.wikipedia.org/wiki/On_the_Internet,_nobody_knows_you%27re_a_dog). The same goes for people who run websites. Ma.gnolia [ma.gnolia.com/] ran an online bookmark service essentially out of someone’s basement on a PowerMac, and failed to institute a solid backup procedure and in the end lost every user’s bookmarks. For bookmarks, perhaps nothing of value was lost, but if Zotero were to suffer a similar fate I think many users would be in outrage.

In fact, Zotero stores your data locally rather than in the cloud, and Zotero publishes guidelines for making backups of your data (http://www.zotero.org/support/zotero_data). I think this example highlights the dangers: how can you be sure your work is safe if it’s hard to even know where it’s stored?

From Keith Krieger: I would suggest that the use of distributed computing resources to only process data is somewhat restrictive.

*Anything* that could be handled with a thin client, from tagging photos with Flickr to manipulating those photos with Adobe’s online version of Photoshop constitutes cloud computing. Or how about adding a Twitter post with a cell phone? Not much computing involved, but very much using the cloud of distributed data processing to interact with other users and data.

And the issue of security seems to arise when considering data in the cloud. We somehow forget about the 1000s of laptops lost or stolen each year, and the breaches of corporate and government data stores that have exposed many SSNs and credit card numbers. Data that is *not* in the cloud seems to be at as much risk as data in the cloud.

As for data loss, I used to be a ma.gnolia user. However, I think that was a disclosure problem rather than a data loss problem. If Larry Haff(Halff? sources differ) had disclosed that ma.gnolia was running on a computer in his basement, I’m thinking that people might have used the service with a different level of expectation, and backed up their data appropriately.

Then again, how many users back up their data appropriately to begin with? Ask people if they have a backup of their digital camera images. The numbers aren’t encouraging.