Queries Response is faster in 6.4 than 6.2

In general having relevant index will improve the query performance. If there is any query that traverse among a large number of nodes then certainly the performance will negatively be affected. Now I am not sure which query you are referring to as being faster in AEM 6.4 but having better indexing can be a potential reason.

In addition to what Hamid mentioned, Indexes are constantly improved from one version to another as Oak version is upgraded with every AEM upgrade. Also, there are new optimizations introduced at oak level that are not available for older versions of oak.

That said, try the following to find more details:

- First, find the query that is slow

- Use the explain query tool to check which index is used for that query on both 6.2 and 6.4

- It might be that different indexes are used for the same query in 6.2/6.4 Or the index definition has been optimized in 6.4

[1] http://<host>:<port>/libs/granite/operations/content/diagnosis/tool.html/granite_queryperformance

Index health plays a vital role in Query performance. There were some enhancements in AEM 6.4 in the field of indexes as described below.

- OAK Run Tool : The Oak Run tool has been added with a new indexing mode, which you can use to handle almost all indexing use cases, without having

to move to JMX. You can re-index your content using this tool. Apart from that this tool gives you out of band support for re-indexing. That means,

if you have limitations on your running production environment and you want to re-index, then this tool will allow you to do the same in a cloned

environment, to avoid critical performance impact.

Oak run tool also allows you to find out the corrupted index, as well as get the report on the indexing statistics for your further analysis.

- Index Health Check addition : A few new sets of health checks have been added around Indexing, Querying and maintenance.

For an instance, Large Lucene Indexes check, which uses Lucene Index Statistics MBean to identify the large indexes, has been added to eventually

control Lucene binaries churn, with the help of a brand-new Lucene Binaries Maintenance feature. Unlike the previous dependency on the successful

data store garbage collection, Lucene binaries can now be cleaned daily basis.

Of course these enhancements doesn't correspond to any change in indexes but it helps to identify the index health and also help in

performing root cause analysis if a query performance is impacted.