On 05/12/2007, ttmdev <[EMAIL PROTECTED]> wrote: > > Given that Marc's producers will be sending non-persistent messages, wouldn't > a shared - as opposed to pure - master/slave configuration provide > redundancy at the broker level and do so w/no extra overhead?
Shared File System & Shared Database only work for persistent messaging and is pretty slow. Pure Master Slave adds considerable latency to each message send on a producer due to waiting for it to be received by a slave. > My thinking > was that if the master were to fail, you can its clients failover to the > slave and thus not burden the other master brokers with the extra message > load. Given the high performance requirements and non-persistence of the messages, just using client side fail-over sounds a more suitable approach. Otherwise it'll take a huge number of producers to reach 200,000 messages/second. Unless you can use async sends on the client I guess - given the non-persistence nature of the messaging? -- James ------- http://macstrac.blogspot.com/ Open Source Integration http://open.iona.com
