Hi,
I am a fourth year undergrad student at the Indian Institute of Technology
(BHU), Varanasi. I am interested in implementing the
caching filesystem idea. Before I submit my proposal I would like to
clarify a few things.
- What are the specific features expected for the administrative
interface, like increasing/decreasing cache size, purging entries before a
particular date, etc. Also would this require a web interface or a command
line interface?
- Similarly what are the required features to be supported for the query
interface.
I thought of implementing it a hash table based LRU cache. There are a few
ways of implementing the in memory cache, which of these methods would be
preferable:
1. Implement the cache within the fuse program, using some form of
thread safe shared memory allocation. Concurrency is now handled by
FUSE. Then the API for querying/admin/statistics would be accessing some
files in the virtual filesystem. The virtual file paths would be used as
URLs.
2. Use an external cache like Redis or Tokyo Tyrant if it doesn't
introduce much overhead.
3. Implement a separate caching server with a listening socket and
handle concurrency by using threads. The API can be implemented as text
commands send to the server socket.
Thanks,
Vivek
_______________________________________________
Monkey mailing list
[email protected]
http://lists.monkey-project.com/listinfo/monkey
