Couch and the Competition

Couch is hardly an original idea. I got the idea for the project after studying up on the Google File System (GFS). There are half a dozen or more distributed file systems in existence anyway, some of them open source. So why building something that's not new?

The biggest reason is because I want to know how to write something that is highly reliable running on shitty hardware. I believe there is a big future in this direction, more and more enterprise applications are spread across multiple machines, and it makes economic sense to make those machines as cheap as possible if the software can tolerate it. So there will be an increasing demand for highly reliable and available software systems on failure prone hardware.

Below are some of the already existing clustered, general purpose file servers that are available:

• IBM's GPFS
• Panasas ActiveScale File System
• IBRIX Fusion
• ADIC StorNext File System
• Red Hat GFS
• Lustre
• Sun ZFS

Of these, only the Red Hat GFS and Lustre systems are open source, although Sun claims the ZFS will be open source as well. All these have a strong emphasis on performance and reliability, but none seem to emphasize low end hardware. Except for ADIC StorNext, None of these are designed to run on or with Windows.

I didn't include Google FS, because it's not available outside of Google. Plus I believe it is not general purpose file system, but is specialized for Google and requires applications to use a special API.

I believe Couch can be successful in a niche. The niche I'm targeting for Couch will be Windows environments that need cheap, reliable storage. They don't care about getting super high transfer rates, they only care that it works reliably, is easy to install and is cheap. There are, I'm sure, lots of companies that meet these criteria and already have all the hardware lying about anyway. They just need the software to put it together.

So I'm not going for super performance or tons of tweakable settings to milk out all possible performance. I want Couch to be easy to install and configure, require no special programming, and be reliable reliable reliable. If it meet thosecriteria, I think it will find a successful niche.

Of course I'm just guessing about the potential demand for Couch and how to target it, what I don't know about all this stuff could fill a book. If anyone has any thoughts on this I'd love to hear them.