Boris Zuckerman. Scalable storage in the cloud.

Cloud Computing Conference - Cloud Slam 2009.
Abstract.
The main reason why anyone should use cloud computing paradigm is scalability.
People want to move large scale grid computations that they used to run on traditional clusters into centrally managed environment, pay for use and be done with it.
Note: these are the same applications that they are already running! Nobody wants to re-write them.
What is the biggest problem of large scale computing? It’s sharing of data. Networks scale easy, cores can be added, but storage traditionally was the bottleneck. You can place huge expensive EMC, HDS, IBM system in the middle of your enterprise and connect it with very expensive FC cables and adapters to your computers, but still it’s very limited. Ibrix has an answer for this. Ibrix allows creating very large scale shared file systems out of inexpensive components.
The Ibrix file system is an absolutely normal, fast Linux and Windows file system. It can be extended on demand by adding storage segments and the servers controlling them. Even more – you do not need designated servers! You can use local computer storage as the part of common file system.
Just imagine: when you rent your cloud space you can choose the parameters of the shared file system. You want very fast? Sure! We have SSD based 200Gb segments, grouped 4 per computing unit. How many do you need? SSDs are too expensive?
Sure! Here is a little cheaper HDS or EMC base segments with designated servers.
Are you planning to live in the cloud for a long time and need affordable large scale storage? Great! Here are large segments based on DELL or HP boxes. How many do you need?

Many Ibrix customers have implemented this model!

Video of the session: