Welcome to the new Gigaspaces XAP forum. To recover your account, please follow these instructions.

Ask Your Question

Space access by multiple PU's in a GSC

Is there any locality advantage if a PU is accessing a non-embedded space if both are located within the same GSC? Or does any space access still have to go through the same remoting protocol as if they were on different servers?

If accessing a space in a different PU is different from accessing an embedded space, what if an application needs multiple PU's (e.g. a Polling container and a Notify container), something like the following:

image description

(this is the general case, I know that I could optimize by embedding the space in either of the 2 PU's). What's the performance of this deployment scenario vs a single PU with an embedded space?

Conversely, if accessing a "local" space (running in the same GSC) is optimized, do we have any means to keep the different PU's co-resident? Particularly if we're using elastic PUs?

asked 2014-11-05 18:11:48 -0600

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted

Firstly, it appears that you are concerned with spaces that have a single partition, otherwise merely locating two different PUs in the same GSC wouldn't guarantee that data lived inside that particular VMs heap. As regards going through a remoting protocol, the answer is both 'yes' and 'no'. There are operations, specifically those of the Change API, that don't require serialiazation and/or marshalling, and thus will not have such a penalty. Nevertheless there will be a network socket for this operation, albeit a local one.

As regards multiple PUs and your example, why can't they both live inside the space? There is nothing that precludes putting more than one such bean inside the same space. In general we frown on co-residency, for reasons involving class loaders and maintainability (changing something in one PU would require both of them to be stopped - but it would be the same if both beans/PUs were in the same heap/GSC).

Hope this helps

answered 2014-11-07 15:50:52 -0600

jb gravatar image
edit flag offensive delete link more

Ah! As I was considering your response, I suddently realized that I was conflating the polling container with the enclosing Processing Unit. Not sure where that came from.

Meaning that I should be able to do this, with as many event containers as needed, yes?

image description

Thanks for your help!

answered 2014-11-07 18:03:00 -0600

edit flag offensive delete link more


As many as needed? If you're going to exceed 100s you should consider using session-based messaging:


jb gravatar imagejb ( 2014-11-18 16:34:24 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower


Asked: 2014-11-05 18:08:07 -0600

Seen: 329 times

Last updated: Nov 07 '14