Why install a non-central Advanced Adapter Engine

[Former Member]

Hi all -

I'm seeing in the PI 7.1 Master Guide that you might want to install a non-central Advanced Adapter Engine for performance reasons.  Anyone have any further information as to what particular reasons those might be to compel a standalone AAE?  -- Geography, throughput, strain on CPU/memory, etc?

Thanks for the help -

Abby

[Former Member]

As far as my understanding goes, non-central AAE will help to locally process the messages in a complex landscape having multiple geographies having different locations - so if it has only a single central instance it will put load on network traffic, single point of failure(unless HA is there) and all load put on a central PI system - but in non-central AAE, locally the msgs can be processed via AAE (if scenario allows bypassing IE) to not even require to reach central PI system having IE also.

So geography, network load, throughput, memory/CPU usage etc all these factors govern the decision of non-central AAE installation in comparison to a single central complete installation of PI system.

[iaki_vila]

Hi Rajeev,

Interesting thread. I'd like to add more elements to this theme. In a complex landscape with DMZ or a proxy is better to put a PCK for company's partners. The PCK has the same architecture that AAE and runs only in Java. As far i know the reason to put a PCK is different that an AAE, it's nearest to a complex infraestructure than a throughput improvement.

Regards

[Former Member]

u can refer below links which points out some of the reasons for installation of de-central adapter engine:

http://www.saptechnical.com/Tutorials/XI/J2EEAdapterEngine/Part1.htm

indeed there is performance improvement by using NC-AAE it provides u a luxury of transferring certain scenarions (using java stack adapters) completly to decentral AAE which eventually results into less burden on central AAE, lesser resource consumption  etc

[baskar_gopalakrishnan2]

In 7.1 There is no option to install one engine(Just AAE alone) as that of 7.3.  You need to install both engines by default. Adapter engine is called here as Advanced Adapter Engine. The main reason is we can simply ignore the integration engine for mesg processing.

The reason for using AAE:  By using one engine, you can simply avoid multiple persistence steps that involved during pipeline processing. Currently classical engine processing( Integration engine) does message processing using java stack and abap stack. So messages are started from sender engine and flowing into integration engine and after processing again go back to receiver adapter (java) engine finally. Also during mapping, the engine reference mapping runtime (java stack) in the middle. since mesg are flowing back and forth between two engines, every time persisted. This can be fully avoided if we go for single engine (AAE) processing. So the performance is pretty much improved by at least 60 percent.

Java engine after jvm 1.6: PI 7.1 uses (Java 1.5+) and so memory garbage collection process is extremely good. so the engine performance is very much improved too. You can monitor the messages fully in RWB.

Refer this document.