| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| Asymmetric performance for send and receive over network |
« View previous topic :: View next topic » |
| Author |
Message
|
| mchan |
 Posted: Tue Nov 01, 2011 9:27 pm Post subject: Asymmetric performance for send and receive over network |
 |
|
Apprentice
Joined: 09 Sep 2011
Posts: 29
|
We are designing for the MQ communication for remote client.
Several communication paths are setup and tested for performance (message throughput and latency)
1 - Send with Direct MQI connection
Remote client [Sender] -> QM (our host network) -> local Q -> Server [Receiver]
2 - Send with HUB and SPOKE
Remote client [Sender] -> Remote Q -> SPOKE QM (client local network) -> HUB QM (our host network) -> local Q -> Server [Receiver]
3 - Receive with Direct MQI connection
Remote client [Receiver] <- QM (our host network) <- local Q <- Server [Sender]
4 - Receive with HUB and SPOKE
Remote client [Receiver] <- Local Q <- HUB QM (client local network) <- SPOKE QM (our host network) <- Remote Q <- Server [Sender]
We found that throughput of case 2 are better than case 1, which is as expected because message channel communication is optimized and better than MQI communication over network.
However
- the throughout of case 3 is quite poor, which is only ~30% of what is in case 1.
- and throughout of case 4 is also poor, which is smaller than case 3.
Could you offer some insight on this?
1) why throughput for receiving message over network is far poorer that sending message
2) why Message channel communication is not better than MQI over network for receiving message
P.S. we have checked with network team and found that there is no difference for send and receive path between our remote client and our host.
Thanks! |
|
| Back to top |
|
 |
| zpat |
| Posted: Wed Nov 02, 2011 1:05 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
For the client app, check out the MQI improvements in WMQ v7, do with read ahead.
| Quote: |
To enable read ahead:
•To configure read ahead at the queue level set the queue attribute, DEFREADA to YES.
•To configure read ahead at the application level:
◦to use read ahead wherever possible use the MQOO_READ_AHEAD option on the MQOPEN function call. It will not be possible for the client application to use read ahead if the DEFREADA queue attribute has been set to DISABLED.
◦to use read ahead only when read ahead is enabled on a queue, use the MQOO_READ_AHEAD_AS_Q_DEF option on the MQOPEN function call. |
Are these messages persistent or non-persistent? Syncpointed or not? |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Nov 02, 2011 5:24 am Post subject: Re: Asymmetric performance for send and receive over network |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| mchan wrote: |
...which is as expected because message channel communication is optimized and better than MQI communication over network.
|
I'm not sure exactly what you are comparing here. It is WMQ Client channel communications across a network (not cross-memory) vs. Sender-Receiver type channel communications across a network (not cross-memory).
Network flows (excluding WMQ from the calculation) will reflect the quality of your network - including stops/hops at firewalls, routers, re-transmissions of packets, etc.) should be identical.
Are the two ends of the channel (either type) identical platforms? Equally provisioned? If not, the smaller, slower platform will certainly exhibit slower throughput - including TCP/IP flows.
So, exactly what (which portions of the throughput) are you monitoring? Just TCP/IP flows? Total round-trip transaction-time (requesting app to receiving app to requesting app)?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| zpat |
| Posted: Wed Nov 09, 2011 4:05 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
| It's impolite to ask a question, and then not reply to those who try to help. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Nov 09, 2011 6:23 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
Could it be a computer science class assignment?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| 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
|
|
|
|
