Please login or register to access secure site features.

Note: By continuing to use DevConnect Program Services you agree to our latest Registered Member Terms.

Sign in using DevConnect ID

Forgot password?

Trouble logging in?

Submit a ticket for Registration Support.

I have an SSO ID

?
sign in

Don't have a DevConnect or SSO ID ?

Create a DevConnect account or join the program.

register now
^
Forum Index » DMCC APIs » Agent state polling and PSN28   XML
 
Author Message
nicklynch



Joined: 26/03/2013 12:28:40
Messages: 8
Offline

We have a product that has to poll the states of 100s to 1000 agents using DMCC's GetAgentState request. We are trying to figure out how best to tune the polling parameters to accommodate higher latency connections (150ms and up).

According to PSN28 ( https://downloads.avaya.com/css/P8/documents/101036842 ), we should limit the number of outstanding messages (messages that have been sent but have not received a reply) to a conservative value - and they give an example of 10 for that limit. Are there any other details on what would be considered conservative? For example, would 100 be fine? Less than 100?

Because of the outstanding message limit and the latency, the messages get backed up in a queue. We are aiming to poll for agent state changes every 2 seconds, but it's taking 15 seconds or more (eg. for 1000 users with a 150ms latency and 10 outstanding, it would be (1000 * 150) / 10 = 15000ms to go through all of the agents) because of these issues.

Increasing the outstanding limit will decrease this wait time, but we're not sure what values would be considered safe.
MartinFlynn



Joined: 30/11/2009 05:00:18
Messages: 1751
Offline

The document is an attempt to help developers avoid overloading an Aura system , especially during transitory events. The numbers shown are just examples and indeed the document describes 10 as an initial, conservative, threshold.

If latency is due to overloading of Communication Manager, it is important to throttle requests. However, if latency is due to the network and Communication Manager is not being overloaded, it may be appropriate to use a larger threshold. Devconnect cannot advise you at to what would be an appropriate threshold for a particular install.

All we can say is that it is important for an application to be able to throttle back requests when the system becomes overloaded and it would be a good idea to design an application in such a way that a craftsperson could tune the application to match the environment.

Please note that there is also an absolute limit on how many messages can be sent to/from Communication Manager per second. See the "System capacities of the AE Services server" section of "Avaya Aura Application Enablement Services Overview and Specification". However, I don't think what you describe would reach this limit.

Martin
 
 
Go to: