Re: Latency issues
Dan Creswell <[email protected]> Mon, 19 Feb 2007 10:49:14 +0000
| Newsgroups | gmane.comp.java.sun.javaspaces |
|---|---|
| Message-ID | <[email protected]> |
asaf.lahav wrote: > Hi all, > > > > I am new to JavaSpaces… so bare with me J > > Let's say I have 100 computers sharing the same Java-Space. > > 2 computers are putting new objects in, and 98 computers are collecting > the objects. > > > > 1. Are the objects inserted into java-spaces stored on multiple > computers (for redundancy)? That's an implementation detail, here's why: JavaSpaces is a specification of a Jini service with a defined set of operations with requirements and effects stipulated for those operations. i.e. It's entirely theoretical though it contains practical considerations for implementation which brings us to: Various implementations of the JavaSpaces spec exist, some run across multiple machines providing replication and/or HA some do not. > 2. If so, Does java-spaces guarantee that only one of the computers > collecting the objects will get a specific object? If the computers collecting objects use take() then the answer is yes but that's because the spec says it must be this way and thus _all_ implementations have to provide this behaviour. > 3. Are all of the computers sharing the same java-space aware of all > the objects inserted into the java-space? I'm not sure what "aware" means in this context. If you are asking "does each computer have equal opportunity for accessing an object" the answer is yes. Again, this is part of the specified behaviour. > 4. Are the objects replicated to all of the computers sharing the > same java-space? > This question is a little unclear to me. You might be asking "if the JavaSpace is implemented as a co-operative service across multiple computers, are the objects replicated to all of these computers?" If that's the case, the answer is "it's implementation dependent". In such arrangements with more than a few computers providing the JavaSpace service, replicating the objects across all members is both inefficient and overkill. Simple implementations may choose to do this, advanced ones may not. Hope that helps, Dan. =========================================================================== To unsubscribe, send email to [email protected] and include in the body of the message "signoff JAVASPACES-USERS". For general help, send email to [email protected] and include in the body of the message "help". To view past JAVASPACES-USERS postings, please see: http://archives.java.sun.com/archives/javaspaces-users.html