Tuning the Dgraph process cache

In the Endeca Server, caching of results for the Dgraph process is used extensively, for many types of queries. You can use the following tips for tuning it.

One of the patterns of caching utilized by the Endeca Server Dgraph process involves caching of intermediate results during parallel query processing. Such a caching strategy allows the computing resources to be utilized better, and therefore, improves throughput of the Endeca Server.

Use the following recommendations when tuning performance of your applications running on top of the Endeca Server:
To specify a new Dgraph cache size for the data domain, use this command:
endeca-cmd put-dd-profile
<data_domain_profile_name> --compute-cache-size <value>

It creates a new data domain profile with the specified setting for the cache size. (In order to have a usable data domain profile, specify other options for the new data domain profile, to mimic an existing data domain profile. These options are not included in this example.) In the new data domain profile, all Dgraph processes will use the specified Dgraph cache size. The existing data domain would need to be recreated utilizing this profile.

Alternatively, instead of using endeca-cmd, you can set the cache size using the putDataDomainProfile operation of the Cluster Web Service, with a value for computeCacheSizeMB, as in this abbreviated example:
<ns1:putDataDomainProfile xmlns:ns1="http://www.endeca.com/endeca-server/cluster/1/0">
  <ns1:dataDomainProfile>
  ...
  <ns2:computeCacheSizeMB xmlns:ns2="http://www.endeca.com/endeca-server/types/1/0">1</ns2:computeCacheSizeMB>
</ns1:dataDomainProfile>
</ns1:putDataDomainProfile>

For additional information on endeca-cmd, see this guide. For full information on managing data domains, see the Oracle Endeca Server Cluster Guide.