|
osdir.com mailing list archive |
|
Subject: Re: Deliver to one only - msg#00051List: java.javagroups.generalSorry, let me explain a little further. The channel holds details about _all_ the members in a group. The DT holds information about what types of services those members provide. The group may be 10 in size but only 2 provide the link to the telemetry system, for example. I see I have thought that under the hood, all messages go to all members anyway so I could have just send the first castMessage to all members anyway. What I want to achieve is the same as in JMS where a message is delivered to just one subscriber. Otherwise the message going on to the telemetry system may be transmitted multiple times and the telemetry system would see this as an escalating condition as if the measured error condition was hit more than once. You might ask why I'm not using JMS and the simple answer is that it doesn't guarantee that the messages is delivered only once across all subscribers when the JMS is not federated. SwiftMQ used to be used and this did work as we wanted but the license changed and the costs for our type of installation are now prohibitive. JGroups, on the other hand, does work well across LANs and WANs. It was my thinking that as JMS is going to be layered on top of JGroups then there may be an interface already within JGroups that supports the notion of delivering a single message to only one member rather than to all. Now you may say I've gone over the top and all I need to do is to create a new group which is the telemetry transmitter group and hey presto everything is fine. The problem is that a single message is still delivered to all members of the group and there are potentially 100s if not 1000s of little services which would mean creating this number of groups and I think that would cause problems although I have not tried this out. IMO you can do this, the way to do it is to pick one member from either the membership, or the info you maintain (and replicate) across the cluster, and send a unicast message to it. But I suggest you use a flow-control based stack rather than the default stack. Hmm, well if you only use unicasts, default stack should do. Have a look at LargeStateTest, I use it to test transferring large states. But I haven't tested 600M yet... -- Bela Ban http://www.jgroups.org Cell: (408) 316-4459 ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click Thread at a glance:
Previous Message by Date: click to view message previewPractical use of vsync.xmlHi, from Bela's description of the vsync protocol stack I found it useful for my needs. However, in my use case the group members are typically switched on pretty much at the same time. Subgroups show up and require merging. But by default a merge protocol is not part of vsync.xml. Why? It works only without merging, if the group members join the group one after another. Is that the intended use case for vsync groups? (or should it work without merge and I'm doing something wrong??) I added MERGE2, but got different problems now. It seems, that the group can't find a consensus about the view. Regards, klaus. ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click Next Message by Date: click to view message previewRe: JGroups evaluation paperHi Emmanuel, I'm looking forward to reading your paper. I'm currently in the process of running my own perftests (org.jgroups.tests.perf), and will soon publish the numbers I got there. Compared to your paper (which I haven't read yet), I'm mostly interested in 1-n (or 2-n) UDP performance, less in n-n. Preliminary results (on Compaq Proliant 2-CPU boxes with 100Mbps switched T) showed 9.2 MB/sec throughput; this almost saturated the network. Even netperf showed on 10.5 MB/sec max. Emmanuel Cecchet wrote: Hi all, Just to let you know that we have submitted for publication a paper evaluating JGroups. The paper is available on the ObjectWeb JMOB Web site at http://jmob.objectweb.org/JGroups.html. All feedback is welcome, Emmanuel et al. -- Bela Ban http://www.jgroups.org Cell: (408) 316-4459 ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click Previous Message by Thread: click to view message previewRe: Deliver to one onlyAndy Hutchinson wrote: I don't know the address and I don't care which of the many gets the message as long as one does. It's standard failover stuff. Actually, Ovidiu has implemented a first prototype of ServerlessJMS on JGroups, which provides this functionality. I suggest you have a look at it (jboss.org). Care to comment Ovidiu ? Cheers, -- Bela Ban http://www.jgroups.org Cell: (408) 316-4459 ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click Next Message by Thread: click to view message previewEncryption problemIs there any example, how to use encrypted connection in multicast communication?? (any exemplary source code) Thanks in advance. Loading Comments...
|
|
