Dynamic Load Balancing

Question

We have developed Client Server Communication through Tuxedo Server and Workstation Client. This approach should handle 500 concurrent requests.
We want to configure our Tuxedo Server in such a way that it could handle all the Requests in parallel.
Currently below is our UBB Configuration
WSL SRVGRP="WSLGRP" SRVID=1 RESTART=Y MAXGEN=255 GRACE=3600
##CLOPT="-A -- -n //10.223.26.28:12791 -m 5 -M 10 -x 15 -p 16914 -P 16924"
##CLOPT="-A -- -n //10.223.26.28:12791 -m 2 -M 4 -x 45" ##Minimum 90 ports needs to be available
CLOPT="-A -- -n //10.223.26.28:12791 -m 2 -M 9 -x 60" ##Minimum 120 ports needs to be available
"REQ_HANDLER" SRVGRP="DATA" SRVID=22280
CLOPT="-A -r -e /log/timings/stderr -o /log/tuxlog/req_handler_stdout"
RQPERM=0666 RPPERM=0666 MIN=1 MAX=200 CONV=N
SYSTEM_ACCESS=FASTPATH MAXGEN=255 GRACE=3600 RESTART=Y
The Problem with the above configuration is that, it is not booting another instance of REQ_HANDLER Server when multiple Request receives.
For the Alternative, we have set both MIN and MAX as 10 to make sure it can handle concurrent request.
But We want the Configuration like if there are multiple request received, it will automatically boot the Another instance of the Server and down it(except MIN value) in case of no request.
Please advise.
Thanks in advance

Answer 1

Hi Vikas,
It's not really possible to give you an answer just by looking at the config; you will need to observe the queue behaviour to get an idea of what is happening; but perhaps the following will help.
You mention that two servers have booted, so since MIN=1 that does indicate that automatic spawning has taken place. What I think is happening is that the two servers are managing with the workload. When you say you "push 20 requests" how quickly did these arrive, and were the servers able to cope with the workload? You can use the pq command in tmadmin to view the queue "REQ_HANDLER" - based on your config the queue would need to exceed the 50 load for at least 5 seconds to cause another server to start, and you may find that by the time the second server has started up the 20 messages have already been processed (I think I am correct in saying that automatic spawning will only be tested again once the second server has completed it's startup and is able to process messages). Also remember that if the queue depth drops below the level you specified, even briefly, the timer starts again from zero. Do you know how long it takes for the service to execute (you can use the -r CLOPT option and run "txrpt" to determine the execution time)? You can also use the psc and psr commands of tmadmin to see which servers are actually processing the messages.
I notice you have used the "L" option in -p, and are working with service loads rather than number of messages. If the messages all have the same load value then you can simply use the number of messages as an alternative in the -p option.
Regards,
Malcolm.