YCSB/hbase1

HBase (1.y) Driver for YCSB

This driver is a binding for the YCSB facilities to operate against a HBase 1 cluster, using a shaded client that tries to avoid leaking third party libraries.

Testing HBase

1. Start a HBase Server

You need to start a single node or a cluster to point the client at. Please see Apache HBase Reference Guide for more details and instructions.

2. Set up YCSB

Download the latest YCSB file. Follow the instructions.

3. Create a HBase table for testing

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', 'family', {SPLITS => (1..n_splits).map {|i| "user#{1000+i*(9999-1000)/n_splits}"}}

Failing to do so will cause all writes to initially target a single region server.

4. Run the Workload

Before you can actually run the workload, you need to "load" the data first.

You should specify a HBase config directory(or any other directory containing your hbase-site.xml) and a table name and a column family(-cp is used to set java classpath and -p is used to set various properties).

bin/ycsb load hbase1 -P workloads/workloada -cp /HBASE-HOME-DIR/conf -p table=usertable -p columnfamily=family

Then, you can run the workload:

bin/ycsb run hbase1 -P workloads/workloada -cp /HBASE-HOME-DIR/conf -p table=usertable -p columnfamily=family

Please see the general instructions in the doc folder if you are not sure how it all works. You can apply additional properties (as seen in the next section) like this:

bin/ycsb run hbase1 -P workloads/workloada -cp /HBASE-HOME-DIR/conf -p table=usertable -p columnfamily=family -p clientbuffering=true

Configuration Options

Following options can be configurable using -p.

• columnfamily: The HBase column family to target.
• debug : If true, debugging logs are activated. The default is false.
• hbase.usepagefilter : If true, HBase PageFilters are used to limit the number of records consumed in a scan operation. The default is true.
• principal: If testing need to be done against a secure HBase cluster using Kerberos Keytab, this property can be used to pass the principal in the keytab file.
• keytab: The Kerberos keytab file name and location can be passed through this property.
• clientbuffering: Whether or not to use client side buffering and batching of write operations. This can significantly improve performance and defaults to true.
• writebuffersize: The maximum amount, in bytes, of data to buffer on the client side before a flush is forced. The default is 12MB. Only used when clientbuffering is true.
• 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.

Additional HBase settings should be provided in the hbase-site.xml file located in your /HBASE-HOME-DIR/conf directory. Typically this will be /etc/hbase/conf.

Bigtable

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.

1. Setup a Bigtable Cluster

Login to the Google Cloud Console and follow the Creating Cluster steps. Make a note of your cluster name, zone and project ID.

2. Launch the Bigtable Shell

From the Cloud Console, launch a shell and follow the Quickstart up to step 4 where you launch the HBase shell.

3. Create a Table

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``.

4. Download the Bigtable Client Jar with required dependencies:

mvn -N dependency:copy -Dartifact=com.google.cloud.bigtable:bigtable-hbase-1.x-hadoop:1.0.0 -DoutputDirectory=target/bigtable-deps
mvn -N dependency:copy -Dartifact=io.dropwizard.metrics:metrics-core:3.1.2 -DoutputDirectory=target/bigtable-deps

Download the latest bigtable-hbase-1.x-hadoop jar from Maven to your host.

5. Download JSON Credentials

Follow these instructions for Generating a JSON key and save it to your host.

6. Create or Edit hbase-site.xml

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_x.BigtableConnection</value>
  </property>
  <property>
    <name>google.bigtable.project.id</name>
    <value>[YOUR-PROJECT-ID]</value>
  </property>
  <property>
    <name>google.bigtable.instance.id</name>
    <value>[YOUR-INSTANCE-ID]</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 have an existing HBase config directory, make sure to add it to the class path via -cp <PATH_TO_BIGTABLE_JAR>:<CONF_DIR>.

7. Execute a Workload

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 hbase1 -p columnfamily=cf -cp 'target/bigtable-deps/*' -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 hbase1 -p columnfamily=cf -cp 'target/bigtable-deps/* -P workloads/workloada