Re: Cleints and workload balancing
"option 1 is fine except that you appear to have 2 redundant boxes.." My thought process was that 2 seperate boxes is safer. If one gets coffee spilled on it and goes down, #2 is still there. Peter Potkay IBM MQSeries Certified Specialist, Developer [EMAIL PROTECTED] X 77906 -Original Message- From: Brian S. Crabtree [mailto:[EMAIL PROTECTED]] Sent: Wednesday, July 03, 2002 10:34 AM To: [EMAIL PROTECTED] Subject: Re: Cleints and workload balancing Peter The question is what are you trying to achieve ? - if it is just the throughput then option 1 is fine except that you appear to have 2 redundant boxes - I would prefer an MQI design rather than using clients but using clients is cheaper. If you want resilience as well then you need at least 2 QMs in front of the processing MQ cluster all in a supercluster so that if either of the front end QMs are not available you can still get messages processed. You can morph this back into Option1 using clients instead of a cluster. Brian S. Crabtree EAI Consultant - Original Message - From: "Potkay, Peter M (PLC, IT)" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, July 03, 2002 9:29 AM Subject: Cleints and workload balancing > 3 machines are set up identically with the same MQ application (get the > request/send back the reply). The goal is to make sure that messages are > processed as fast as possible. > > QM1 is a spoke in our Hub and Spoke architecture. QM1 sends the original > request to QMHub. From QMHub, the message goes to QM2, our client > concentrator. If I set up the three application machines as MQClients, and > have them all open the SAME request queue with an unlimited wait, what will > happen? I assume one of the three (random?) will grab the 1st request and > process it. If another request message lands before whoever grabbed the > first can come back to do its get with unlimited wait again, then the > remaining 2 will fight for the next message. If the all 3 are there waiting, > then any of them, and possibly always the same one, will process the next > message. > > So while I don't have true work load balancing (at the end of the day > Client1 may have got 23%, Client2, 57% and 20% the remaining, and the next > day the numbers could be different), no messages were ever waiting on the > request queue (unless all 3 were busy at the same time). Any pitfalls here? > > Or do I need to forget the QM2 Client concentrator and go to QM1 ---> QMHub > > Cluster1, where I install a QM on each of the 3 machines, cluster the > three and then have a cluster queue called RequestQ on all 3. Will messages > leaving QMHub (not in Cluster1) land in a round robin fashion on the 3 > instances of RequestQ? > > > Peter Potkay > IBM MQSeries Certified Specialist, Developer > [EMAIL PROTECTED] > X 77906 > > > > This communication, including attachments, is for the exclusive use of > addressee and may contain proprietary, confidential or privileged > information. If you are not the intended recipient, any use, copying, > disclosure, dissemination or distribution is strictly prohibited. If > you are not the intended recipient, please notify the sender > immediately by return email and delete this communication and destroy all copies. > > Instructions for managing your mailing list subscription are provided in > the Listserv General Users Guide available at http://www.lsoft.com > Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Re: Cleints and workload balancing
n the second scenario, why not move the clustered queue up into QM1 and forgo the hub? The Hub houses MQSI, which is used for this request/reply to transform Cobol copyboks to XML and back. Peter Potkay IBM MQSeries Certified Specialist, Developer [EMAIL PROTECTED] X 77906 -Original Message- From: Rick Tsujimoto [mailto:[EMAIL PROTECTED]] Sent: Wednesday, July 03, 2002 10:56 AM To: [EMAIL PROTECTED] Subject: Re: Cleints and workload balancing In the first scenario, if the overall throughput is satisfactory, what difference does it make that you didn't achieve "true" load-balancing? If load-balancing is important you could always write a program for QM2 to redistribute the messages equally. The program would be easy to write and you still benefit by using free clients. In the second scenario, why not move the clustered queue up into QM1 and forgo the hub? If the hub is just a pass-thru, I don't see the point of going to it just to abide to an architecture, especially if performance is an issue. You could make QM1 a cluster unto itself. This approach costs more, especially if the client boxes have 2 or more processors. "Potkay, Peter M (PLC, IT)" To: [EMAIL PROTECTED] Subject: Cleints and workload balancing Sent by: MQSeries List 07/03/2002 09:29 AM Please respond to MQSeries List 3 machines are set up identically with the same MQ application (get the request/send back the reply). The goal is to make sure that messages are processed as fast as possible. QM1 is a spoke in our Hub and Spoke architecture. QM1 sends the original request to QMHub. From QMHub, the message goes to QM2, our client concentrator. If I set up the three application machines as MQClients, and have them all open the SAME request queue with an unlimited wait, what will happen? I assume one of the three (random?) will grab the 1st request and process it. If another request message lands before whoever grabbed the first can come back to do its get with unlimited wait again, then the remaining 2 will fight for the next message. If the all 3 are there waiting, then any of them, and possibly always the same one, will process the next message. So while I don't have true work load balancing (at the end of the day Client1 may have got 23%, Client2, 57% and 20% the remaining, and the next day the numbers could be different), no messages were ever waiting on the request queue (unless all 3 were busy at the same time). Any pitfalls here? Or do I need to forget the QM2 Client concentrator and go to QM1 ---> QMHub > Cluster1, where I install a QM on each of the 3 machines, cluster the three and then have a cluster queue called RequestQ on all 3. Will messages leaving QMHub (not in Cluster1) land in a round robin fashion on the 3 instances of RequestQ? Peter Potkay IBM MQSeries Certified Specialist, Developer [EMAIL PROTECTED] X 77906 This communication, including attachments, is for the exclusive use of addressee and may contain proprietary, confidential or privileged information. If you are not the intended recipient, any use, copying, disclosure, dissemination or distribution is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return email and delete this communication and destroy all copies. Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Re: Cleints and workload balancing
In the first scenario, if the overall throughput is satisfactory, what difference does it make that you didn't achieve "true" load-balancing? If load-balancing is important you could always write a program for QM2 to redistribute the messages equally. The program would be easy to write and you still benefit by using free clients. In the second scenario, why not move the clustered queue up into QM1 and forgo the hub? If the hub is just a pass-thru, I don't see the point of going to it just to abide to an architecture, especially if performance is an issue. You could make QM1 a cluster unto itself. This approach costs more, especially if the client boxes have 2 or more processors. "Potkay, Peter M (PLC, IT)" To: [EMAIL PROTECTED] Subject: Cleints and workload balancing Sent by: MQSeries List 07/03/2002 09:29 AM Please respond to MQSeries List 3 machines are set up identically with the same MQ application (get the request/send back the reply). The goal is to make sure that messages are processed as fast as possible. QM1 is a spoke in our Hub and Spoke architecture. QM1 sends the original request to QMHub. From QMHub, the message goes to QM2, our client concentrator. If I set up the three application machines as MQClients, and have them all open the SAME request queue with an unlimited wait, what will happen? I assume one of the three (random?) will grab the 1st request and process it. If another request message lands before whoever grabbed the first can come back to do its get with unlimited wait again, then the remaining 2 will fight for the next message. If the all 3 are there waiting, then any of them, and possibly always the same one, will process the next message. So while I don't have true work load balancing (at the end of the day Client1 may have got 23%, Client2, 57% and 20% the remaining, and the next day the numbers could be different), no messages were ever waiting on the request queue (unless all 3 were busy at the same time). Any pitfalls here? Or do I need to forget the QM2 Client concentrator and go to QM1 ---> QMHub > Cluster1, where I install a QM on each of the 3 machines, cluster the three and then have a cluster queue called RequestQ on all 3. Will messages leaving QMHub (not in Cluster1) land in a round robin fashion on the 3 instances of RequestQ? Peter Potkay IBM MQSeries Certified Specialist, Developer [EMAIL PROTECTED] X 77906 This communication, including attachments, is for the exclusive use of addressee and may contain proprietary, confidential or privileged information. If you are not the intended recipient, any use, copying, disclosure, dissemination or distribution is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return email and delete this communication and destroy all copies. Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Re: Cleints and workload balancing
Peter The question is what are you trying to achieve ? - if it is just the throughput then option 1 is fine except that you appear to have 2 redundant boxes - I would prefer an MQI design rather than using clients but using clients is cheaper. If you want resilience as well then you need at least 2 QMs in front of the processing MQ cluster all in a supercluster so that if either of the front end QMs are not available you can still get messages processed. You can morph this back into Option1 using clients instead of a cluster. Brian S. Crabtree EAI Consultant - Original Message - From: "Potkay, Peter M (PLC, IT)" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, July 03, 2002 9:29 AM Subject: Cleints and workload balancing > 3 machines are set up identically with the same MQ application (get the > request/send back the reply). The goal is to make sure that messages are > processed as fast as possible. > > QM1 is a spoke in our Hub and Spoke architecture. QM1 sends the original > request to QMHub. From QMHub, the message goes to QM2, our client > concentrator. If I set up the three application machines as MQClients, and > have them all open the SAME request queue with an unlimited wait, what will > happen? I assume one of the three (random?) will grab the 1st request and > process it. If another request message lands before whoever grabbed the > first can come back to do its get with unlimited wait again, then the > remaining 2 will fight for the next message. If the all 3 are there waiting, > then any of them, and possibly always the same one, will process the next > message. > > So while I don't have true work load balancing (at the end of the day > Client1 may have got 23%, Client2, 57% and 20% the remaining, and the next > day the numbers could be different), no messages were ever waiting on the > request queue (unless all 3 were busy at the same time). Any pitfalls here? > > Or do I need to forget the QM2 Client concentrator and go to QM1 ---> QMHub > > Cluster1, where I install a QM on each of the 3 machines, cluster the > three and then have a cluster queue called RequestQ on all 3. Will messages > leaving QMHub (not in Cluster1) land in a round robin fashion on the 3 > instances of RequestQ? > > > Peter Potkay > IBM MQSeries Certified Specialist, Developer > [EMAIL PROTECTED] > X 77906 > > > > This communication, including attachments, is for the exclusive use of > addressee and may contain proprietary, confidential or privileged > information. If you are not the intended recipient, any use, copying, > disclosure, dissemination or distribution is strictly prohibited. If > you are not the intended recipient, please notify the sender > immediately by return email and delete this communication and destroy all copies. > > Instructions for managing your mailing list subscription are provided in > the Listserv General Users Guide available at http://www.lsoft.com > Archive: http://vm.akh-wien.ac.at/MQSeries.archive Instructions for managing your mailing list subscription are provided in the Listserv General Users Guide available at http://www.lsoft.com Archive: http://vm.akh-wien.ac.at/MQSeries.archive