First the java code has to be compiled:
mvn package
The script run.sh can then be used to run a series of tests, currently:
- a full suite of document tests as baseline using 1 thread and 3 shards
- document, vertex, edge, and AQL tests with 8, 32, and 64 threads and 3 shards, and
- document, vertex, edge, and AQL tests with 1 thread and 3, 9, 27, and 81 shards.
the results are output as CSVs into a directory with name javabench-$(date --iso-8601=minutes)
(so for example javabench-2019-06-26T13:12+00:00), and plotted using gnuplot (which needs
to be installed for this script to run through).
The following parameters are accepted:
# ./run.sh --help
Usage:
-e|--endpoint ArangoDB endpoint
-r|--requests number of requests
-s|--shards comma-separated list of shard numbers to test
-t|--threads comma-separated list of thread numbers to test
For the 3.5 release candidates the script was run as follows
# ./run.sh -e sup4.arangodb.org:8529
Where sup4.arangodb.org was one computer in a ArangoDB cluster started (using
the arangodb starter) on set of 3 scaleway machines.
These tests run for more than a day. To have a more concise testing run you can decrease the number of requests.
The runscript runs arangodb-java-benchmark.jar.
Note that you have to run an insert benchmark before you can run any of the others, i.e. to run
document_update you first have to run document_insert.
java -jar arangodb-java-benchmark.jar --test document_insert
It is not necessary to run vertex_insert for the edge tests.
--acquireHostList <Boolean> automatic acquire list of endpoints to use for load balancing
(default: false)
--batchSize <Integer> number of operations in one batch
(necessary only when API supports batching) (default: 1)
--collection <String> collection name to use in test (default: ArangoJavaBenchmark)
--connections <Integer> number of parallel connections per thread (default: 1)
--database <String> database name to use in test (default: ArangoJavaBenchmark)
--delay <Integer> delay (in seconds) to use between runs
(necessary only when --runs > 1) (default: 0)
--docArraysSize <Integer> size of array fields in the documents (default: 10)
--docIndexArrays <Collection> comma separated list of types of indexes on array
fields. possible values: "hash", "skiplist",
"persistent", "geo", "fulltext"
--docIndexLargeSimple <Collection> comma separated list of types of indexes on large
String fields. possible values: "hash", "skiplist",
"persistent", "geo", "fulltext"
--docIndexObjects <Collection> comma separated list of types of indexes on nested
objects. possible values: "hash", "skiplist",
"persistent", "geo", "fulltext"
--docIndexSimple <Collection> comma separated list of types of indexes on String
fields. possible values: "hash", "skiplist",
"persistent", "geo", "fulltext"
--docLargeSimpleSize <Integer> size of large String fields in the documents (default: 100)
--docNestingDepth <Integer> max depth of nested objects in the documents (default: 1)
--docNumArrays <Integer> number of array fields in the documents (default: 0)
--docNumIndexArrays <Integer> number of array fields to be indexed (default: 1)
--docNumIndexLargeSimple <Integer> number of large String fields to be indexed (default: 1)
--docNumIndexObjects <Integer> number of nexted objects to be indexed (default: 1)
--docNumIndexSimple <Integer> number of String fields to be indexed (default: 1)
--docNumLargeSimple <Integer> number of large String fields in the documents (default: 0)
--docNumObjects <Integer> number of nested objects in the documents (default: 0)
--docNumSimple <Integer> number of String fields in the documents (default: 5)
--docSimpleSize <Integer> size of String fields in the documents (default: 20)
--dropDB <Boolean> drop DB before run (default: false)
--duration <Integer> number of seconds the test should run
(if > 0 the option 'requests' is ignored) (default: 0)
-e,--endpoints <String> comma separated list of endpoints to connect to
(default: 127.0.0.1:8529)
--edgeCollection <String> edge collection name to use in graph (default:
ArangoJavaBenchmarkEdge)
--graph <String> graph name to use in test (default: ArangoJavaBenchmarkGraph)
--keyPrefix <String> document key prefix (necessary only when run multiple times)
--loadBalancing <LoadBalancingStrategy> load balancing strategy to use (for cluster setup).
possible values: "none", "round_robin", "one_random"
(default: none)
--numberOfShards <Integer> number of shards of created collections (default: 1)
--outputInterval <Integer> output interval in seconds (default: 1)
-p,--password <String> password to use when connecting.
--protocol <Protocol> network protocol to use. possible values: "vst",
"http_json", "http_vpack" (default: vst)
--query <String> Custom AQL query (supported bind params: @@collection,
@@vertex, @@edge, @graph, @doc, @docs, @key, @@keys)
--replicationFactor <Integer> replication factor of created collections (default: 1)
--requests <Integer> number of operations per thread (default: 1000)
--runs <Integer> run test n times. drop database between runs (default: 1)
-t,--test <Collection> comma separeted list of test cases to use.
possible values: "version", "document_get",
"document_insert", "document_import",
"document_update", "document_replace",
"aql_custom", "aql_get", "aql_insert", "aql_replace",
"vertex_get", "vertex_insert", "vertex_update",
"vertex_replace", "edge_get", "edge_insert",
"edge_update", "edge_replace"
--threads <Integer> number of parallel client threads (default: 1)
-u,--user <String> username to use when connecting (default: root)
--vertexCollection <String> vertex collection name to use in graph (default:
ArangoJavaBenchmarkVertex)
--waitForSync <Boolean> use waitForSync for created collections (default: false)