Ambient Data Filter Performance Issue

Question

We are experiencing page load times between 5-15 seconds (sometimes more) when applying a small load of approximately 40 users to our web application. This is for a simple page with just one static on it. Investigation uncovered that the Content service is consuming approximately 50% of our server CPU which maxes out the CPU and hence leads to the long page loads.

Removing the “SDL Ambient Client Filter” from our web.xml reduces page load to within 200 milliseconds, and reduces the Content service CPU to just 1% (even with 100+ active users).

We have pin-pointed the issue to the SDL Ambient Client Filter which is making two calls to the Content service for every page load: /client/v4/content.svc/RequestEndActionImport /client/v4/content.svc/RequestStartActionImport

The offending code is in the class “com.sdl.web.ambient.client.AmbientClientFilter” inside the “content-ambient-client-*.jar”. Specifically the following lines:

Builder actionBuilder = new Builder().withReturnType(String.class).withActionName(isRequestStart ? "RequestStartActionImport" : "RequestEndActionImport").withActionParameter("ClaimStore", String.format("\"%s\"", new Object[]{SERIALIZATION_UTIL.serializeClaimStore(claimStore)}));

    . . .

String claimStoreStr = StringEscapeUtils.unescapeHtml4((String) ContentClientProvider.getInstance().getContentClient().performAction(actionBuilder.build(), false));

All other calls to the Content service are cached on the webapp level (as configured in the cd_client_conf.xml), hence the huge performance improvement without the ADF enabled.

When we looked into the calls above, and we found that they are returning some claim values, and claim scopes (claims like useragent:browser, os:model, tracking:id, and session:id). We're not sure if this is necessary or if there is some way to avoid this large overhead when using the ADF.

Here is the filter as configured in our Web.xml:

    <filter>
        <filter-name>SDL Ambient Client Filter</filter-name>
        <filter-class>com.sdl.web.ambient.client.AmbientClientFilter</filter-class>
    </filter>
    <filter-mapping>
       <filter-name>SDL Ambient Client Filter</filter-name>
       <url-pattern>/*</url-pattern>
    </filter-mapping>

Here are some useful implementation details:

Language = Java   
CM = SDL Web 8.5  
WebApp CIL = com.sdl.web -> cil -> 8.5.0-1056  
Microservices = Hotfix CD_8.5.0.15928  
Webserver = WebSphere Custom
Implementation = CIL: Not DXA, Not DD4T
Add ons: SmartTarget with custom cartridges

Any ideas on why these calls are occurring for every page request or if there is some way to cache/disable them?

Edit: Note that we are using SmartTarget actively on some pages.

Answer 1

Answer to your question, yes, Indeed It's expected and executed when a new session starts, onRequestStart at the beginning of each request, onRequestEnd at the ending of each request if ADF enabled, yes it can disable.

In Java web app web.xml comment out this section to disable Ambient Data Framework if you use CIL.

<filter>
    <filter-name>Ambient Client Filter</filter-name>
    <filter-class>com.sdl.web.ambient.client.AmbientClientFilter</filter-class>
</filter>
<filter-mapping>
    <filter-name>Ambient Client Filter</filter-name>
        <url-pattern>/*</url-pattern>
</filter-mapping>

Ambient Data Framework (ADF) - It's basically a common store of contextual data about the current request in your web application. Your Web app and SDL products all read and write data to this store to communicate with each other and drive contextual experiences.

ADF works as a repository of information related to a specific session or to a specific request that can be accessed or updated during a web operation.

If you are not using any of these below features and not using ADF yourself in your implementation, it is recommended to disable it.

• Session Preview (XPM)
• Experience Optimization
• Audience Manager User
• Generated Content
• Context Expressions

To learn more about ADF what it's? and how it works? in detail then refer to these following blogs and explained well.

Ambient Data Framework in a Nutshell

Who’s who in the Ambient Data Framework

I hope it helps.