Best architecture for highest throughput with failover and reliability in place

[Postol <ft...@centrum.cz>]

Hello All,

I want to ask which architecture should I use for the following scenario:

I have a server, which will be consuming messages from a couple of
publishers (lets say 10). These publishers will post messages to separate
queues. Number of messages will be about 400 per second at peak from each
system.

When server receives the message, it will process it and processed message
will publish to connected clients. Lets say that the number of clients will
not be higher than 50.

Requirments for the architecture is failover, reliability (no duplicates, no
missing messages) and performance! :) I know that failover + reliability is
against performance, but I need all of these three.

So the question is, how should I design the system for the highest possible
throughput with failover and reliablity in place?


When I am publishing from server to client, which type of destination should
I use?

1) Topic with durable subscribers
2) X queues, where X is number of connected clients
3) Virtual topic with X consumers

Which version is the best one for given scenario?

Is it possible to get a throughput of activemq broker, so the server will
publish about 4000 messages per second at peaks and all clients will be able
to receive them imediatelly?

Thanks in advance for any suggestion.

Postol
-- 
View this message in context: http://www.nabble.com/Best-architecture-for-highest-throughput-with-failover-and-reliability-in-place-tp22562821p22562821.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.