Oracle recommends the following practices to ensure high performance and high availability.
Use the Dgraph in multithreaded mode and experiment with increasing the number of threads. By default, the Dgraph runs in multithreaded mode with the number of threads set to one. It can be configured to run with a larger number of threads.
Protect your configuration from hardware failures:
Use redundant disk drives with RAID (RAID 0, RAID 0+1, RAID 5), or SAN.
Utilize device redundancy for servers, load balancer devices and routers.
Use a second data center (hot or cold standby) to protect from site failures, such as power and network outages or enterprise data center failures.
Protect your configuration from software failures:
If you have not done this already, upgrade to 64-bit systems to avoid “out of address space” failures. (Starting with its version 6.*, MDEX Engine installations are only supported on 64-bit systems.)
Use respawning monitors to protect against unexpected fatal process errors.
Watch out for paging with process memory usage.
Periodically examine your application for slow queries, or massive responses (too many results returned not all of which may be needed by the users).