[chord] About: an application based on Chord protocol, ADB or BerkeleyDB

[Tri Nguyen Phi Minh]

Dear Dr.Emil Sit,

We compiled Chord-0.1-20070418 version with sfslite 0.8.16.
We successfully tested a tool, named "filestore", in some Linux systems.

Our group would like to develop an application based on Chord protocol.
Here is our target in nutshell:
- In our system, each computer connects to some sensor devices, they're
used for medical purpose. It gathers sensor data and parses them into the
form of pairs of <key, value>. Whenever a pair of <key, value> is gathered,
then we will use Chord for distributing them over the network. For storing
them permanently, we'd like to use a key-value storage engine (i.e.,
LevelDB or BerkeleyDB).

We are reading the mailing list and CFS source code since we think we need
to modify it to fit our requirement. It is complicated, so I post my
question with hope you could help us clarify a bit.
  1. Could you please give us hints which layers of CFS  we should modify?
I guess we could start with your filestore tool, I think it breaks a file
into blocks, names them with keys and then distributes them. It sounds
similar to our need.

  2. I observed  that CFS supports ADB and BerkeleyDB in implementation.
Could you tell me in which case we use ADB instead of BerkeleyDB? What are
advantages for choosing ADB or BerkeleyDB. Besides, you mentioned DB3 in
the post in 2004 (
http://pdos.csail.mit.edu/pipermail/chord/2004-August/000292.html), is it
another DB that CFS supports to store blocks? Actually, we are thinking
about using LevelDB instead of them.


I really appreciate your help.
Thank you very much.
Tri Nguyen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://amsterdam.lcs.mit.edu/pipermail/chord/attachments/20121206/f3e5a01e/attachment.html