ALA applications do not currently have expected latencies as performance requirements, to judge whether changes are going to have a negative impact on users.
Each of the core applications should have their latencies measured over time (see Uptime records) and expected latencies set to sufficiently low values that the applications feel responsive.
The worst case currently for ALA seems to be the Occurrence record view that is being served from a single-node Cassandra database (which Cassandra is not designed for and which the JVM Heap is currently at 200% of its maximum recommended value), at between 1.5 and 5 seconds.
ALA applications do not currently have expected latencies as performance requirements, to judge whether changes are going to have a negative impact on users.
Each of the core applications should have their latencies measured over time (see Uptime records) and expected latencies set to sufficiently low values that the applications feel responsive.
The worst case currently for ALA seems to be the Occurrence record view that is being served from a single-node Cassandra database (which Cassandra is not designed for and which the JVM Heap is currently at 200% of its maximum recommended value), at between 1.5 and 5 seconds.