Hello Chris:

I think you are basically right and is very likely to be that way for a while. However I think that SSDs can give a completely turning point in the not too far future. SSDs are by far faster (several factors) than any magnetic media in both reading and writing and could comunicate with the computer by optical means. There is also tecnology coming to create very huge flash drives (around 1 TB of data). On the other side if things keep the way they are and it doesn't changes towards that even with the pressure they are being pushed forward, a centralized huge RAID system over high speed lan is very likely to cost less in both maintenance and deployment with probably very similar results if not better and no software is required at all wich means less possible bugs or costs in training personal both in the user side and the sysadmin side. Also there is the added benefict of redundancy for the data protection side. I think it still has a market but not the way you are envisioning it. The idea is certainly not new. What you are talking about is in fact a P2P system or a distributed filesystem. Is not that I like to discourage anybody from building stuff but maybe you find my comments usefull and might save you some time. Anyway I would find insteresting ideas or comments from others that prove me wrong in any case is in any case an interesting discussion.

Regards
Waldo

Waldo,

A few clarifications related to your comment.

First, I don't mean to imply that Dispersed Storage is a competitive alternative to a solid state drive (SSD) -- they are very complementary. I actually believe that the emergence of solid state drives in portable devices (phones, laptops, etc.) enables a better storage architecture than the local hard drive model in your PC architecture. Specifically, a local SSD in a portable device with a dsNet providing secondary / backup / archive storage provides better performance, security and capabilities than a local-hard-drive-based system.

(I agree that you'll eventually see 1 TB SSDs, but the cost per unit storage for hard drives will remain substantially better for the foreseeable future.)

One fundamental problem of a RAID system is that the reliability of RAID decreases as the size of the systems grows. RAID can be configured so you can loose either one storage node -- typically a hard drive -- with (one-dimensional parity) or two drives in a stripe with two-dimensional parity. If you just have 10 hard drives of data to store, the odds are good that you won't loose more than two drives at once, so RAID works well. However, if you have 1,000, 10,000 or 100,000 hard drives of data, then the odds get very high that you'll loose more than two drives at once, so a single RAID array wouldn't provide high data reliability. So, you end up having to replicate that RAID array once or more in a large system to maintain high reliability. The extra overhead of RAID parity times multiple RAID array copies equals lots of excess storage. And this rate of "storage overhead" increases as the size of the storage system grows. So, a centralized huge RAID system gets less efficient as it gets bigger.

In addition, a single large RAID array is vulnerable to a location failure.

A Dispersed Storage network, like a packet switched network, gets more efficient as it gets bigger. So the larger the storage system, the better Dispersed Storage works compared to RAID.

I should also point out that a dsNet retrieves and stores data like a P2P system; however, P2P systems traditionally use replicated files or replicated blocks of data. A Dispersed Storage Network does not store files nor source data nor blocks of source data. The "slices" stored on a dsNet are a mathematical transformation of the source data which have engineered properties for security, privacy, reliability and availability. (And slices are also not compressed nor encrypted data -- they are different.)

A dsNet virtualizes the data in that it appears to save and retrieve data, but it doesn't store actual data.