Invitation to TBVFS Virtual Gathering

Invitation to TBVFS Virtual Gathering

[Herbert Poetzl]

For a while now I've been thinking about adding a feature
to Linux, which might be useful in many different areas, but,
and that's the reason why I'm spending my time on this, seems
to come especially handy for the creation and maintenance of
template based virtual servers.

The idea of a Template Based Virtual Filesystem is not new to
Linux or other Operating Systems, as a matter of fact it pops
up in news and mailing list since 1993 under various names,
nevertheless it seems that none of these Linux Implementations
was able to survive more than a few years or has proven worthy
to be included in the Linus kernel tree.

I would very much like to see all the knowledge and findings
of the past ten years combined in one place to form a conclusive
document (or howto) for creating such a feature or filesystem,
so we can avoid all the nasty pitfalls and drawbacks former
implementations carried.

Therefore, I'd like to invite you to join in and share your ideas,
opinions and experiences. Any input and/or help will be valued
and welcomed. 

best,
Herbert 

PS: please forward, if you know someone who would be interested
    in participating ...

Re: Invitation to TBVFS Virtual Gathering

[Erez Zadok]

In message <20021226012244.GA22897@www.13thfloor.at>, Herbert Poetzl writes:
> 
> For a while now I've been thinking about adding a feature
> to Linux, which might be useful in many different areas, but,
> and that's the reason why I'm spending my time on this, seems
> to come especially handy for the creation and maintenance of
> template based virtual servers.
> 
> The idea of a Template Based Virtual Filesystem is not new to
> Linux or other Operating Systems, as a matter of fact it pops
> up in news and mailing list since 1993 under various names,
> nevertheless it seems that none of these Linux Implementations
> was able to survive more than a few years or has proven worthy
> to be included in the Linus kernel tree.
> 
> I would very much like to see all the knowledge and findings
> of the past ten years combined in one place to form a conclusive
> document (or howto) for creating such a feature or filesystem,
> so we can avoid all the nasty pitfalls and drawbacks former
> implementations carried.
> 
> Therefore, I'd like to invite you to join in and share your ideas,
> opinions and experiences. Any input and/or help will be valued
> and welcomed. 
> 
> best,
> Herbert 
> 
> PS: please forward, if you know someone who would be interested
>     in participating ...

I assume you've heard of my own stackable templates work, named FiST.  It is
still ongoing, with plenty of people developing new stackable file systems
and applications based on my templates (both commercially and academically).
You can get more info from: http://www.cs.columbia.edu/~ezk/research/fist/.

Coincidentally, we just released a new version of the fistgen package.

Erez.

Re: Invitation to TBVFS Virtual Gathering

[David Chow]

>
>
>>Linux or other Operating Systems, as a matter of fact it pops
>>up in news and mailing list since 1993 under various names,
>>nevertheless it seems that none of these Linux Implementations
>>was able to survive more than a few years or has proven worthy
>>to be included in the Linus kernel tree.
>>
>>I would very much like to see all the knowledge and findings
>>of the past ten years combined in one place to form a conclusive
>>document (or howto) for creating such a feature or filesystem,
>>so we can avoid all the nasty pitfalls and drawbacks former
>>implementations carried.
>>
>>Therefore, I'd like to invite you to join in and share your ideas,
>>opinions and experiences. Any input and/or help will be valued
>>and welcomed. 
>>
>>best,
>>Herbert 
>>
>>PS: please forward, if you know someone who would be interested
>>    in participating ...
>>    
>>
>
>I assume you've heard of my own stackable templates work, named FiST.  It is
>still ongoing, with plenty of people developing new stackable file systems
>and applications based on my templates (both commercially and academically).
>You can get more info from: http://www.cs.columbia.edu/~ezk/research/fist/.
>
>Coincidentally, we just released a new version of the fistgen package.
>
>Erez.
>
We made template based template-based physical hosts instead of virtual 
hosts (althongh virtual servers will still work wih user mode linux). We 
made stackable file systems for implementing templates for Linux hosts, 
I refer to "host" not servers because Linux hosts not necessary servers. 
As we are using this technology to deploy and manage desktops, cluster 
nodes and servers, or whatever you want to call it. We are well 
experience in implementing file system templates for different Linux 
distributions such as Redhat, SuSE, Mandrake and even the latest 
UnitedLinux . My experience tells me there are no standard way in 
implementing file system based templates for all Linux distribution ( I 
assume you want a VFS based template implementation as you are posting 
this message to the fsdevel list). I start with FiST (Erez's project) as 
a research study couple of years ago. We finally implement our own file 
systems and the project is quite sucessful and now commericialized as a 
software product (see http://www.shaolinmicro.com) . But there are still 
lot of issues on VFS based implementation such as dcache validation, 
locks and all sorts of cache consistency issues.  Since we are in a 
multiple physical hosts situation rather than virtual, so that we used 
NFS for host-to-host fs linking. There are really infinite number of 
unknown issues to solve as I am sure you want something work pratically 
rather than just doing some hacking to the Linux kernel. The realistic 
is that Linux kernel runs with a Linux distribution which changes 
everyday and differ from every vendors. Not sure are you looking for 
this but this is what we've done in the past. You can get more info from 
our website http://www.shaolinmicro.com or download the software for 
some hack.

regards,
David