This driver is a binding for the YCSB facilities to operate against a HBase 1.0.x Server cluster or Google's hosted Bigtable.
To run against an HBase 0.94.x cluster, use the hbase094 binding.
To run against an HBase 0.98.x cluster, use the hbase098 binding.
See hbase098/README.md for a quickstart to setup HBase for load testing and common configuration details.
In addition to those options available for the hbase098 binding, the following options are available for the hbase10 binding:
durability: Whether or not writes should be appended to the WAL. Bypassing the WAL can improve throughput but data cannot be recovered in the event of a crash. The default is true.
Google's Bigtable service provides an implementation of the HBase API for migrating existing applications. Users can perform load tests against Bigtable using this binding.
Login to the Google Cloud Console and follow the Creating Cluster steps. Make a note of your cluster name, zone and project ID.
From the Cloud Console, launch a shell and follow the Quickstart up to step 4 where you launch the HBase shell.
For best results, use the pre-splitting strategy recommended in HBASE-4163:
hbase(main):001:0> n_splits = 200 # HBase recommends (10 * number of regionservers)
hbase(main):002:0> create 'usertable', 'cf', {SPLITS => (1..n_splits).map {|i| "user#{1000+i*(9999-1000)/n_splits}"}}
Make a note of the column family, in this example it's `cf``.
The Bigtable protocol uses HTTP/2 which requires an ALPN protocol negotiation implementation. On JVM instantiation the implementation must be loaded before attempting to connect to the cluster. If you're using Java 7 or 8, use this Jetty Version Table to determine the version appropriate for your JVM. (ALPN is included in JDK 9+). Download the proper jar from Maven somewhere on your system.
Download one of the bigtable-hbase-1.# jars from Maven to your host.
Follow these instructions for Generating a JSON key and save it to your host.
If you have an existing HBase configuration directory with an hbase-site.xml file, edit the file as per below. If not, create a directory called conf under the hbase10 directory. Create a file in the conf directory named hbase-site.xml. Provide the following settings in the XML file, making sure to replace the bracketed examples with the proper values from your Cloud console.
<configuration>
<property>
<name>hbase.client.connection.impl</name>
<value>com.google.cloud.bigtable.hbase1_0.BigtableConnection</value>
</property>
<property>
<name>google.bigtable.cluster.name</name>
<value>[YOUR-CLUSTER-ID]</value>
</property>
<property>
<name>google.bigtable.project.id</name>
<value>[YOUR-PROJECT-ID]</value>
</property>
<property>
<name>google.bigtable.zone.name</name>
<value>[YOUR-ZONE-NAME]</value>
</property>
<property>
<name>google.bigtable.auth.service.account.enable</name>
<value>true</value>
</property>
<property>
<name>google.bigtable.auth.json.keyfile</name>
<value>[PATH-TO-YOUR-KEY-FILE]</value>
</property>
</configuration>
If you wish to try other API implementations (1.1.x or 1.2.x) change the hbase.client.connection.impl appropriately to match the JAR you downloaded.
If you have an existing HBase config directory, make sure to add it to the class path via -cp <PATH_TO_BIGTABLE_JAR>:<CONF_DIR>.
Switch to the root of the YCSB repo and choose the workload you want to run and load it first. With the CLI you must provide the column family, cluster properties and the ALPN jar to load.
bin/ycsb load hbase10 -p columnfamily=cf -cp <PATH_TO_BIGTABLE_JAR> -jvm-args='-Xbootclasspath/p:<PATH_TO_ALPN_JAR>' -P workloads/workloada
The load step only executes inserts into the datastore. After loading data, run the same workload to mix reads with writes.
bin/ycsb run hbase10 -p columnfamily=cf -jvm-args='-Xbootclasspath/p:<PATH_TO_ALPN_JAR>' -P workloads/workloada