| Author |
Message
|
| pcelari |
 Posted: Mon Jan 14, 2008 12:19 pm Post subject: sequence of connector in AggrCtrl node possible? |
 |
|
Chevalier
Joined: 31 Mar 2006
Posts: 411
Location: New York
|
this is the first time I use aggregate node. I use it to parellelize a set of requests to services to build a final combined reply.
since some service takes longer than others, I would like to submit such requests first to shorten overall response time. So I need to somehow control the sequence of connectors connected to the AggrCtrl node.
Is there a way to do this?
If not, I hope in the next release IBM will add such capability to the out terminal of AggrCtrl, or just the inputNode's out terminal.
_________________
pcelari
-----------------------------------------
- a master of always being a newbie |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Mon Jan 14, 2008 12:23 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Huh?
You don't need to link your AggregationRequest nodes to your AggregationControl nodes any more, in v6 and later.
You sequence when your requests start by... sequencing when you send the request message that goes to the server (and then sends info to AggregationRequest node).
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| pcelari |
| Posted: Mon Jan 14, 2008 12:48 pm Post subject: |
|
|
Chevalier
Joined: 31 Mar 2006
Posts: 411
Location: New York
|
thanks the the response.
No, I don't link them directly. I connect several compute nodes to AggrCtl node to build different requests and put them into different queues for the different service providers to process, in parellel. Each route ends with an AggrRqst node just to specify the folder name.
I wonder if I can somehow control the sequence with which different routes are executed. I want to build the request that takes the longest to process first. Is there a way to do this?
thanks,
_________________
pcelari
-----------------------------------------
- a master of always being a newbie |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Mon Jan 14, 2008 12:51 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Okay, so...
You're doing
Compute1->
Compute2->
Compute3->AgCntl->ComputeX->AggRequest
Compute4->
Compute5->
I'm suggesting you do
AgCntl->Compute(Flow)-1->Compute1->AgRequest
-2->Compute2->AgRequest
...
-5->Compute5->AgRequest
Where you use out, out1, out2, etc. on Compute(Flow) to order your requests.
Or just use AgCntl->Compute(propagate)->AgRequest
and build each request in Compute, then propagate, then build next request, and etc.
And then you can control the order as well.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| pcelari |
| Posted: Mon Jan 14, 2008 1:07 pm Post subject: |
|
|
Chevalier
Joined: 31 Mar 2006
Posts: 411
Location: New York
|
thanks for the insight.
I'm doing this:
MQInpt -> AgCtl -> Compute1 -> AggRqst1
-> Compute2 -> AggRqst2
-> Compute3 -> AggRqst3
...
and want to control the sequence between AgCtl and ComputeX.
It seems this approach doesn't make sense as long as there is no way to control the sequence of the routes.
Your approach apparently is the best way to accomplish what I'm trying to do. Many thanks!
However, how can you handle the situation when more than 5 routes need to be connected to the out* terminals of a compute node? Current version has only out, ..., out4.
_________________
pcelari
-----------------------------------------
- a master of always being a newbie |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Mon Jan 14, 2008 1:31 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
There's always PROPAGATE TO LABEL...
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| pcelari |
| Posted: Mon Jan 14, 2008 1:59 pm Post subject: |
|
|
Chevalier
Joined: 31 Mar 2006
Posts: 411
Location: New York
|
woo...ow!
how come that command in the manual escaped my eyes..., until you listed it.
That's the difference between a GM and non-... !
many many thanks,
_________________
pcelari
-----------------------------------------
- a master of always being a newbie |
|
| Back to top |
|
|
|
|
