Project idea regarding transparency for distributed systems to be incorporated in the btrfs design

Project idea regarding transparency for distributed systems to be incorporated in the btrfs design

[sparx, India]

Hi,
We are a group of four undergraduate students studying in final year
completing our computer engineering course from the University of
Pune, India.

We would like to do an operating system related project over a period
of 6 months.
We have studied and understood the design of Btrfs.

We came across this paper presented in Fast '07 on transparent file
systems in distributed networks. (TFS)
The paper basically suggests some design changes in a regular file
system so as to support (and minimize) the efforts of contributory
applications for storage resources in a distributed system. This would
enable greater resource contribution without affecting local
performance ( Storage contribution can increase upto 40% in corporate
networks)

To support transparent memory allocation to a shared pool efficiently
a file sytem should:
1.Allocate most blocks in a given region
2.View storage media as a series of chunks
3.Some replicative power of its own.
However the TFS design does not interfere with a file systems own
allocation policy.

 We could not find any article/information which specifically mentions
that Btrfs handles memory allocation in distributed systems
transparently.

We would like to ask your suggestions on the following points:

1. Are we correct in our assumption that Btrfs does not handle
transparent allocation as yet or have we missed something?
2. Would the Btrfs community appreciate this contribution   to the
file system?(i.e. do you think it is worth our investing our time and
effort in this and that it contributes positively to btrfs?)

Thank-you for your consideration.
-Sparx group

Re: Project idea regarding transparency for distributed systems to be incorporated in the btrfs design

[Chris Mason]

On Sun, Aug 09, 2009 at 12:29:23PM +0530, sparx, India wrote:
> Hi,
> We are a group of four undergraduate students studying in final year
> completing our computer engineering course from the University of
> Pune, India.
> 
> We would like to do an operating system related project over a period
> of 6 months.
> We have studied and understood the design of Btrfs.
> 
> We came across this paper presented in Fast '07 on transparent file
> systems in distributed networks. (TFS)
> The paper basically suggests some design changes in a regular file
> system so as to support (and minimize) the efforts of contributory
> applications for storage resources in a distributed system. This would
> enable greater resource contribution without affecting local
> performance ( Storage contribution can increase upto 40% in corporate
> networks)
> 
> To support transparent memory allocation to a shared pool efficiently
> a file sytem should:
> 1.Allocate most blocks in a given region
> 2.View storage media as a series of chunks
> 3.Some replicative power of its own.
> However the TFS design does not interfere with a file systems own
> allocation policy.
> 
>  We could not find any article/information which specifically mentions
> that Btrfs handles memory allocation in distributed systems
> transparently.
> 
> We would like to ask your suggestions on the following points:
> 
> 1. Are we correct in our assumption that Btrfs does not handle
> transparent allocation as yet or have we missed something?

I'm not sure I fully understand the concept of transparent allocation.
Could you please describet his in more detail?

> 2. Would the Btrfs community appreciate this contribution   to the
> file system?(i.e. do you think it is worth our investing our time and
> effort in this and that it contributes positively to btrfs?)

We're always interested in btrfs contributions ;)  Thanks for taking the
time to read about the FS and propose this.

-chris