Documentation & best practices

TarMK is the default microkernel of choice for AEM. Only select MongoMK or RDBMK if Adobe guidance mandates it. 

RDBMK is under restricted support, only to be used with AEM Forms Author, and limited to the supported RDBMS.

The data store is the part of the repository that stores large binary files, where as the nodestore stores small data like node definitions and properties. It is recommended the nodestore is persisted by the mechanism aligned with the selection microkernel (TarMK stores the nodestore in tar files, MongoMK stores the nodestore in MongoDB) however the data store should always be stored in the file system or an Amazon S3 bucket(s).

File Data Store (FDS)

  • The File Data Store (FDS) is the default data store type for TarMK and MongoMK in AEM.
  • FDS can be persisted on local disk, NAS or SAN
  • FDS is also sharable via the Shared Data Store configuration (see Shared Data Store section below).
  • Any production deployment should have at a minimum a File Data Store configured for all AEM instances (Author and Publish).

Amazon S3 Data Store (S3DS)

  • The Amazon S3 Data Store (S3DS) stores files in Amazon S3 buckets. 
  • S3DS is "infinitely scalable" and recommended for Data Stores over 5TB in size.
  • S3DS is sharable via the Shared Data Store confguration (see below).

Shared Data Store

The Shared Data Store configuration allow a single Data Store to be used by multiple AEM instances. This offers a number of efficiencies for AEM systems where large/volumes of files are shared across AEM instances, but also has several limitations.

Garbage collection is the process of re-claiming disk space form the repositories Data Store. Garbage collection must be run frequently, during off-hours.

Garbage collection on a Shared Data Store

Gargbage collection and TarMK cold standby

AEM features and enhancements

Key repository features

AEM upgrades can be broken out into phases, each phase with a set of preparatory and execution steps. Upgrades to AEM 6.4 require in-place upgrades, rather than side-grades if backward compatibility is required.


Planning the upgrade

Pre-upgrade steps

Understanding the upgrade procedure

Performing the upgrade

Post-upgrade steps

Oak provides two main types of indexes: Lucene Property Indexes and Property Indexes.

SOLR server integration

Re-indexing should only be performed for limited reasons.

Tools and approaches


Writing queries

Understanding and troubleshooting queries

denotes community supported tooling.