Situation: we're got a dev server (4 cores, 16GB memory) that's running a single-instance ArangoDB (3.11.7) and a Spring Boot application that queries ArangoDB via Spring Data 4.1.0. We're posting here instead of in Spring Data as we use the Spring template that passes through to the Java driver directly (see below).

Complication: we've got a query that returns ~6 MB in JSON. That exact query in the admin interface returns in 0.5s. The same query via the Java driver below takes 14s. Regardless of admin interface/ Spring, the CPU/memory barley touches 15% each, same in the Arango dashboard/ metrics, so it's not a spec issue.

watch.start("aql");
ArangoCursor<FindAll> cursor = ops.query(query, bindVars, options, FindAll.class);
watch.stop();

(Note: FindAll is a POJO that includes Arango-annotated classes)

Questions:

1. How can we debug this properly?
2. How can we increase parsing performance? 550 lines should not take 13 seconds, regardless of size.

UPDATE:

1. Setting batchSize to 1 and 1000 - this has shown that the query itself returns in the same time as in the admin console (0.6) for batchSize=1, shifting the 13-second delay to the cursor.asListRemaining() method.
2. Setting 'RawJson' as the return type - 0.8s total (of which 0.6s is the query), so the performance hit is in the conversion, though it's hard to imagine 550 rows (~6 MB) taking 14 seconds.
3. Created duplicates of all Arango-annotated classes without annotations - 14s down to 2s.

==> does anyone have input on how to improve the deserialization to the original annotated classes?