activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tim Bain <tb...@alumni.duke.edu>
Subject Re: Evaluating AMQ architecture capabilities
Date Sat, 30 May 2015 22:20:58 GMT
If you want brokers to resume sending messages when other brokers fail, you
want master/slave pairs (http://activemq.apache.org/masterslave.html);
simply writing the messages to a JDBC database doesn't mean other brokers
will automatically start reading/sending them.

KahaDB is generally considered to be faster than JDBC, and it can be used in
a master/slave configuration as long as the file system to which it writes
is accessible by both brokers in the pair via NFS or something similar.

What's your need for a network of brokers?  Are you sure your needs won't be
met by just a single master/slave pair (where only one of the two brokers is
active at any time)?  And if you do, what network topology and distribution
of clients will you be using, and why is that the appropriate one for your
needs?  Networks of brokers increase complexity and might deliver no
performance improvement if the right topology isn't selected, so make sure
you're not just assuming that a network of brokers is necessary to get the
scalability you need.

Tim


hadrian wrote
> Hi, 
> Am trying to understand more on the tool capabilities; looking for a
> scalable, high available and persistent msg solution.  Am thinking of a
> network of brokers (with store-and-forward to address the limitation of
> brokers with no consumers) each having JDBC persistence.  In case one of
> the brokers goes down would the others be able to retrieve via JDBC the
> messages persisted by the faulty one?  
> Thank you,
> Horea





--
View this message in context: http://activemq.2283324.n4.nabble.com/Evaluating-AMQ-architecture-capabilities-tp4697042p4697109.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.

Mime
View raw message