| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| Clustering for scalability and a FIFO requirement |
« View previous topic :: View next topic » |
| Author |
Message
|
| ivanachukapawn |
 Posted: Tue Aug 22, 2006 5:13 am Post subject: Clustering for scalability and a FIFO requirement |
 |
|
Knight
Joined: 27 Oct 2003
Posts: 561
|
| Application A has a requirement to process literally 10's of millions of MQ messages daily (maybe 40 or 50 million). Clustering is being considered as a possible way to provide the scalability which is obviously going to be needed. The idea would be to have N queue managers (each on its own 4-way Solaris server) each configured with the application's queues Then MQ would load balance the message traffic to the application spread across multiple servers. However, if there is a requirement for this application to process all of these messages in FIFO, can this clustering approach be made to work? If so, how? |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Tue Aug 22, 2006 5:19 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
You can't guarantee FIFO order across queue managers.
Clustering will distribute messages across all instances of a queue. It is entirely likely that message 1 will go to one queue manager, message 2 to another, and etc. with no particular guarantees about order.
40-50 million messages a day? How big? What counts as "processing"? Getting the message off the queue? Sending a response in a certain period of time after the message was sent? Are the messages coming as a lump set, or at regular intervals or at a variable rate? Are they coming from one source, or many sources? Do the messages have to be persistent?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| Vitor |
| Posted: Tue Aug 22, 2006 5:22 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
If you've got N queue managers with N queues with N copies of application A then you're not going to manage FIFO in terms of oldest message first, just most likely the oldest message first.
If you mean by FIFO that the result of processing message 2 depends on message 1 then that's called message affinity. This too will wreck a cluster solution, though I can't imagine all 50 million messages rely on each other in a giant chain.... 
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
