Random thoughts...

[howard chen]

I don't know the implementation details of a traditional DFS, but
seems most DFS is suitable only for read intensive applications, but
not for write intensive use...

Why?

Is it related to replication?

If yes, can we delay the replication a little bit (say. when system
loading is low)

Or even more, can we disable the replication for a particular set of
files, say cache files, we afford to lose the cache, but we want to
take the advantage of using DFS for performance gain. (This is similar
to wikipedia's implementations but we talking for file storage, we
don't need a Berkeley db)




Just some random thoughts...


howa