[tahoe-dev] Tahoe as cloud back end for a local application

Nathan nejucomo at gmail.com
Tue Feb 21 19:56:25 UTC 2012


On Tue, Feb 21, 2012 at 1:34 AM, Eugen Leitl <eugen at leitl.org> wrote:
> On Tue, Feb 21, 2012 at 10:21:59AM +0100, Eugen Leitl wrote:
>>
>> I'm investigating porting an Apache/PHP/MySQL application currently
>> using local storage to use a private Tahoe grid (servers with public
>> IPs) as the backend. Total storage size is about 10 TBytes,
>> using about a million files. The application would typically
>> run on local LAN behind NAT.
>>
>> Question: how many servers would one need to make this reliable?
>> How much storage should each server contribute for good performance?
>> How slow would access to such files be (user queries local
>> DB via a http front end, then retrieves the hits from the cloud),
>> all assuming fast servers on the public Internet? Under 10 seconds,
>> over 10 seconds?
>
> Oh, one thing I forgot: the data in the cloud is to be read-only for
> application users. It is r/w to the administrator group.

I'm quite interested in any answers you discover about performance
within a single enterprise and when integrating with existing stacks.
How do you plan on porting a LAMP app?  I'd love to see a blog post or
article describing your experience, because this desire may be both
common and complex to implement.


I am not certain if this helps, but I've created an nginx proxy
configuration specifically to implement read-only access to Tahoe and
to preserve privacy between clients to the proxy:

https://bitbucket.org/nejucomo/lafs-rpg/


Please be careful if you want to rely on this for security, and please
let me know if you discover any security, configuration, or
performance problems.

I am somewhat new to nginx so I have no claims about performance.  I
believe it may read entire requests before forwarding, and it *may*
read entire responses before forwarding them.  For my use case of a
web server serving several dozen small http files / directories with
infrequent, enabling nginx caching of webapi responses might greatly
reduce latency.


Nejucomo


> _______________________________________________
> tahoe-dev mailing list
> tahoe-dev at tahoe-lafs.org
> http://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev



More information about the tahoe-dev mailing list