Fred Wilson published a blog post today with the premise that there will be no files in the cloud. With due respect I am going to have to object to his premise.

He follows the logic of the Web, which is always a smart path to follow when thinking of how data flows. He argues that some day there will never be a need for downloads.

His conclusion:

So if you are working in the cloud storage space, I think you’ve got a bit of a conundrum. The reality of the market today is that people use files. You need to support that use case, enhance it, and make people’s lives easier. But over time, that use case will go away. And what people will want is a service that doesn’t have files as the atomic unit. And how do you elegantly morph from a file centric model to a document centric model? It won’t be easy, I’m sure of that.

Wilson is right for the most part. The storage world is turning upside down. Traditional storage vendors are offering proprietary solutions. These are big box plays. It’s a bolt-on approach, mechanical solutions that have associated hardware costs, potentially significant total costs of ownership and lots of moving parts such as power supplies, fans and cabling.

Gluster

Last week I interviewed a few of the executives from Gluster, a company that counts Amazon Web Services, GoGrid and Pandora as customers. These companies use Gluster technology for its storage software which allows any file-based application to work inside a cloud-based infrastructure.

Wilson talks about the beauty of Google Docs. He’s right – it is easier to move a link around then it is a file. A Google Doc item is an object. Gluster is a storage system that allows people to store and access data as an object and as a file. It can be easily replicated and distributed across a cluster of inexpensive hardware with ease and simplicity for people to consume, often through a Web oriented environment.

What this all means?

Actually, storage is a hot topic.  Gluster unifies file and block storage. If one of Wilson’s startups so wished, it could create Amazon S3 on EC2 and Elastic Block Storage. There could be back up from the data center to AWS, while simultaneously providing for  S3 functionality in the data center itself.  And again, it all runs on commodity hardware.

Wilson is a music fan. That’s what lead him to writing this post and which brings me to my last point about Pandora.

Pandora uses Gluster due to the very fact that any song may have as many as 12 file formats. The service needed a distributed storage environment for the 1.2 petabytes of audio it serves every week.  The music service now has three data centers, each with a six-node GlusterFS cluster. It replicates for high availability and it has more than 250 terabytes of total capacity. It can scale, requires one person to manage it and no changes are ever needed for the application itself.

Services Angle

Our guess is that it will be quite some time before we are all in the cloud, with no need for file-based cloud storage. More likely is considerable disruption in the proprietary space where EMC, NetApp and the big technology companies play.

Gluster is not for everyone. Still, about 90% of all apps remain file based. There will be files in the cloud for quite some time, especially as the infrastructure delineating public clouds and private data centers continues to be blurred. What we will see is a decreasing reliance on hardware for storage. The demand will be to replace hardware with software, which is cheaper to run and easier to manage.