I'm not an expert on NIO but might a memory mapped file work if all you are saving is blobs?
Sean On 8/31/07, John McClain - Sun Microsystems, Inc. <[EMAIL PROTECTED]> wrote: > Mike Morris wrote: > > John McClain - Sun Microsystems, Inc. wrote: > > > >> The problem with logstore is that it depends on Progress Software's > >> ObjectStore PSE Pro. > > > > Thanks for the clarification, John. > > > >> Because of this I don't see having a dependency on PSE Pro was being > >> an option for River. > > > > (I probably should have peeked at the code before commenting.) In my > > mind was the option of replacing PSE Pro with some other > > (licence-compatible) OODB (not that I have the cycles, energy or > > knowledge to try, right now :-) > > As it turns out for the way Outrigger uses the store OODBs aren't really > a good fit. It really just needs a bunch of independent blobs so in lots > of ways an OODB (and even a RDB) is overkill for Outrigger - so if we > decided snapstore wasn't enough I would probably go in some other > direction. For the record the option the options we have considered in > the past are : > > A JDBC based store : In the past one issue here was being able > to provide a complete integrated package so it works out of the > box without having to bring your own RDBMS. Now that Apache > Derby is around (and its in the JDK!) an embedded Derby could be > the default, and integration of others could be left to the user > > Keeping all the state in the log files snapstore (and logstore) > produce, deleting/compacting logs as entries go away. Probably > the most of work of all the options we have considered, but > minimizes external dependencies and gives us the most options > for optimization (not that it would necessarily lead to the > fastest Outrigger...) > > Berkeley DB Java Edition : seems to be a good fit with what > Outrigger needs, not sure about licensing, it is available > under an "OSI-certified" open source license but I don't > know which license it is and if there are any license > compatibility issues with Apache. > > A "commit to memory" based scheme where Outrigger commits > transactions to memory in one or more back end processes running > on other machines. Has the potential to be very fast, but > isn't as general as the other options. > > I could see ending up with multiple stores to satisfy different sets of > requirements. > > > -- > John McClain [EMAIL PROTECTED] > Sun Microsystems, Inc. > Burlington, MA > > And it is that way today. We are tricked by hope into starting > companies, beginning books, immigrating to this country and investing > in telecom networks. The challenges turn out to be tougher than we > imagined. Our excessive optimism is exposed. New skills are demanded. > But nothing important was ever begun in a prudential frame of mind. > > - David Brooks > > >
