Throttling – allowing for a majority to go Out

[Rob Lindsey]

Ok… I know that this has been hashed and rehashed but I am still not good on my understanding of throttling

Problem:

Inbound thread on process A has been down for 5+ hours and it sends to multiple threads in process B.  Sending system has 30,000 data messages queued.  

Question:

What are the best settings to have to make sure that there is not a large backup in the recovery database (i.e. the most msgs I want queued up in the engine would be a 10% total (in this case 3000))?

Known:

The engine prioritizes inbound data.  Already have inbound process setup at min 200 max 400 and 50%.  This gave us a backup of over 27,000+ in the recovery database.  The inbound thread is able to handle up to 150 msgs  a second.  Now when the inbound stopped sending data it cleared out the data messages in less than 60 minutes.  The bad thing was that the oldest message in the recovery database was almost 5 hours old.

Thanks in advance for any advice, comments, things to think about.

Rob

Topic

[James Cobane]

I don’t know that there are any hard and fast values to set the throttling at to obtain a specific threshold.

[Troy Morton]

Hi Rob,

At a previous employer we had throttling set to 1 / 1 / 100%.  That seemed to keep the state 7 queues from building for the most part.  But, I think this setting might slow down the Xlate process in general.  

Have you tried using the mulit-threaded Xlate option?  Logically, it seems maybe if the Xlate is multi-threaded, it might free up more CPU time for the other threads in the process (protocol thread) to send inter-process data?  Just guessing on that one.

Good luck.  Reply back and let every know if you find something that works!

Troy Morton

[Rob Lindsey]

Thanks guys for some encouragement.

We do have Multi-threaded Xlate option turned on.  

After some experimentation we are trying the following settings   Min: 5 Max: 10 %: 100 to see how that works.  I will have to keep you guys up to date on how we think it will work.  So far it seems to be doing OK but of course we have not had the inbound thread go down for a long period of time and to be honest I do not want to “try” that just to see how it does, to be honest.

[Author]

Replies