Vint Cerf: It's like 1973 for Moving Data Around in the Cloud

Written by Alex Williams / January 8, 2010 11:03 AM / 7 Comments

Using a cloud computing service may sound enticing, but you better consider how that data can be moved around if you want to switch to a different provider.

It's a big problem that now has the attention of Vint Cerf, who is calling for standards to define how customer data gets passed between different cloud service providers.

Cerf, Google's chief Internet evangelist, is one of those legends of the tech world, up there with people like Steve Wozniak. He is one of the co-designers of the TC/IP protocol. He is one of those few who had this idea way back when of hooking computers together to create a network. Today we call that network the Internet.

So you listen when Cerf gets up to speak and says that it's like 1973 out there when it comes to cloud computing data portability.

According to InfoWorld, Cerf said major cloud service providers like Amazon, Google and IBM have no real form of interoperability. Cerf spoke Thursday night at the Churchill Club in Menlo Park, Ca.

"We don't have any inter-cloud standards," Cerf said. "The current cloud situation is similar to the lack of communication and familiarity among computer networks in 1973."

People will want to move data around. They may have multiple cloud service providers. They may want to use different cloud service providers as an interconnected network. Moreso, customers will simply want to move data from Cloud A to Cloud B.

Cerf went on to say that the industry needs to develop protocols and standards to make this all happen. It's important to note that Google, Cerf's employer, obviously has a stake in how this all pans out.

We went to Aardvark to ask about this issue.

What can you do right now to avoid getting locked into one cloud service provider?

Marc Limotte, director of engineering at Feeva Technology, writes:

"The obvious problem is that the difficulty in switching limits consumer choice and therefore competition. You can't "vote with your feet", if you can't walk away.
This is common in IT, though. It's never been easy to switch from one enterprise package to another, or from one hosting facility to another.

The data isn't even the worst of the problem. In most cases, you can at least get an extract (even if it is terabytes of data), and perform a load in to some other system. The more complex issue is when you architect your solution to take advantage of a vendor's proprietary services (e.g. the data store in Google App Engine, or the Amazon's SQS). Not that you shouldn't use these features... they're useful, just be aware that they start to limit your options is you want to someday move away from that platform.

My suggestion... make sure you know how to export your data. And try and use your own interfaces in front of custom services. that way if you want to move, you just have to write an adapter, and not a complete re-architecture."

Comments

Subscribe to comments for this post OR Subscribe to comments for all ReadWriteEnterprise posts

If we're talking about PIM data, wouldn't the de facto standard be SyncML?

Posted by: Open attitude | January 8, 2010 12:33 PM

If you've every tried to develop an Amazon Machine Image (AMI) you know how tricky portability is because there is no way to run it outside of EC2. In particular, you can't run an AMI locally or in your own hosted environment.

Of course, the same applies to Google App Engine (GAE) apps. There is a development simulator that can run locally, but there is no way that I know of to run a GAE app outside of Google. If there was, you could theoretically make an AMI that ran GAE apps.

Anyway, I'd be interested to hear what Cerf thinks about Java. The JVM is a virtual machine after all - it just uses the native operating system as the hypervisor. If people wrote all of their apps for the JVM over standard data access APIs, then we'd have a high degree of portability.

Posted by: josh | January 8, 2010 2:50 PM

While Vint Cerf's comments may make some sense... How about having the best of both worlds - both "Streaming" and "Downloads"; and, after all "VIDEO SEARCH" can be big business!

Posted by: aa batteries | January 10, 2010 1:46 AM

Of course, the same applies to Google App Engine (GAE) apps. There is a development simulator that can run locally, but there is no way that I know of to run a GAE app outside of Google. If there was, you could theoretically make an AMI that ran GAE apps.

Posted by: sesliturkey | January 11, 2010 2:28 AM

Fundamentally, all these cloud services are contrary to the decentralized nature of the Internet. I'm surprised Vint Cerf even works at Google.

Posted by: Sivan | January 11, 2010 12:00 PM

This is a huge problem because it is not about the data, but rather it is about the "features" that each vendor enables. A database is a database, and, Microsoft Access 97 aside, it is pretty simple to export data to any other format. What is very difficult to do is to move the features within any particular implementation. As it is difficult to migrate custom Photoshop Actions/Scripts to The GIMP, so it is difficult to migrate GAE code to any other platform.

This challenge will no doubt require decades of work before a standard, or set of standards, are developed to allow for easy migration. Until then, if you can keep your people from using any of the "features" of the cloud providers that are unique to them, that would be prudent if you want to keep your migration options open.

At least, back in 1973 (just pre- modern email), people had not already settled into their current patterns, so the introduction of a new protocol/standard was less likely to encounter confusion. These days, anything new would need to dive deep into often proprietary programming in order to be effective, so we can expect the development, political grandstanding and implementation times to increase dramatically. Maybe even "decades" is too short ...

Posted by: Stupidscript | January 11, 2010 4:58 PM