Quick Start

This section describes how to run YCSB on OrientDB running locally.

1. Set Up YCSB

Clone the YCSB git repository and compile:

git clone https://github.com/brianfrankcooper/YCSB.git
cd YCSB
mvn clean package

Note: alternatively the latest docker image could be used

docker run -d --name orientdb -p 2424:2424 -p 2480:2480 -e ORIENTDB_ROOT_PASSWORD=admin orientdb:latest

2. Run YCSB

Now you are ready to run! First, load the data:

./bin/ycsb load orientdb -s -P workloads/workloada

Then, run the workload:

./bin/ycsb run orientdb -s -P workloads/workloada

See the next section for the list of configuration parameters for OrientDB.

DB creation with the OrientDBClient

This client will create a database for you if the connection database you specify does not exists. You can also specify connection information to a preexisting database.

You can use the orientdb.newdb=true property to allow this client to drop and create a new database instance during the load phase.

NOTE: understand that using the orientdb.newdb=true property will drop and recreate databases even if it was a preexisting instance.

WARNING: Creating a new database will be done safely with multiple threads on a single YCSB instance, but is not guaranteed to work when launching multiple YCSB instances. In that scenario it is suggested that you create the db before hand, or run the load phase with a single YCSB instance.

OrientDB Configuration Parameters

• orientdb.url - (required) The address to your database.
  • Supported storage types: memory, plocal, remote
  • EX. plocal:/path/to/database
  • remote EX. remote:172.16.50.15:2424/database
• orientdb.user - The user to connect to the database with.
  • Default: admin
• orientdb.password - The password to connect to the database with.
  • Default: admin
• orientdb.server.user - The server user to connect to the database with.
  • Default: root
• orientdb.server.password - The server password to connect to the database with.
  • Default: admin
• orientdb.newdb - Overwrite the database if it already exists.
  • Only effects the load phase.
  • Default: false
• orientdb.remote.storagetype - Storage type of the database on remote server
  • This is only required if using a remote: connection url

Plot results

To plot workload results install gnuplot, then create a directory to collect the outputs

mkdir -p orientdb/outputs/orientdb

load the data

./bin/ycsb load orientdb -s -P workloads/workloadc 

and run the workload, e.g. workloadc, by following the naming convention for the file name outputRun_<workload>.txt:

./bin/ycsb run orientdb -s -P workloads/workloadc > orientdb/outputs/orientdb/outputRun_workloadc.txt

After a successful worload run, goto the plots directory cd orientdb/plots and run

./tool.sh "plot" "save" ../outputs/orientdb/

The diagrams will be in the images directory.

Known Issues (with ODB v2.2.x)

• There is a performance issue around the scan operation. This binding uses OIndex.iterateEntriesMajor() which will return unnecessarily large iterators. This has a performance impact as the recordcount goes up. There are ideas in the works to fix it, track it here: #568.
• Iterator methods needed to perform scans are Unsupported in the OrientDB API for remote database connections and so will return NOT_IMPLEMENTED status if attempted.