Skip to content

Latest commit

 

History

History
118 lines (87 loc) · 6.26 KB

README.md

File metadata and controls

118 lines (87 loc) · 6.26 KB

YCSB

Requirements

  • Installed Java13+ on machines where YCSB will be run.
  • Installed YDB Python SDK. You can install it with command pip3 install ydb numpy requests ydb[yc].
  • Installed parallel-ssh on machine where scripts will be run. We worked with parallel-ssh on version 2.3.4.
  • Requirements for the selected database: YDB, CockroachDB, YugabyteDB

Getting Started

First, you need to deploy one of the databases on the machines. For more details on each database:

Now you can start running the benchmark on the selected database.

Below is the instruction for running YCSB for YDB, CockroachDB, YugabyteDB, but first you need to familiarize yourself with YCSB workload.

Workload

You can read about YCSB workloads here.

We should have two config files to run YCSB: <DB_NAME>.rc and workload.rc.

The workload.rc file contains the configuration for the YCSB workload. The following variables can be found in this file:

  • WORKLOADS - a list of workloads to be executed.
  • RECORD_COUNT - the number of records in the database at the start of the workload.
  • OP_COUNT_TOTAL - the number of operations to be performed.
  • DISTRIBUTIONS - what distribution should be used to select the records to operate on – uniform, zipfian, hotspot, sequential, exponential or latest
  • YCSB_THREADS - the number of YCSB client threads (default - 64).
  • YCSB_THREADS_DE - the number of YCSB client threads for workload D and E (default - 512).
  • LOAD_YCSB_THREADS - the number of YCSB client threads when load the data.
  • KEY_ORDER - should records be inserted in order by key (“ordered”), or in hashed order (“hashed”).
  • MAX_PARTS, MAX_PART_SIZE_MB, LOAD_DATA - these are settings for developers and the default values are suitable in most cases.

YDB

Next, you need to configure the YDB config file according to your needs.

Suppose that on ydb_host1.com YDB is running. We want to run YCSB workload against this database from machines ycsb-host1.com, ycsb-host2.com, etc.

If you look into the config file ydb.rc, you can find:

  • TARGET - one of the hosts where YDB is running (that is, ydb_host1.com).
  • DATABASE_PATH - the path to the database on which the performance tests will be conducted.
  • STATIC_NODE_GRPC_PORT - GRPC port of the static node, by default 2135.
  • YCSB_HOSTS - list of hosts on which YCSB will be run (this is, ycsb-host1.com, ycsb-host2.com, etc).
  • YCSB_HOSTS_COUNT - if you want to limit the number of YCSB_HOSTS without changing the list.
  • YCSB_PATH - path to the folder with YCSB on YCSB_HOSTS.
  • YCSB_TAR_PATH - if YCSB_PATH is not present, the archive at this path will be unpacked in YCSB_DEPLOY_PATH.

After configuring ydb.rc and workload.rc (more about it above), you can start YCSB:

cd <PATH_TO_BENCHHELPERS>/ycsb
./run_workloads.sh --name ycsb-ydb-test-run1 --log-dir <PATH_TO_LOG_DIR> configs/workload.rc configs/ydb.rc

There are also parameters for run_workloads.sh:

  • --name - for convenience, the log file names will be suffixed with <NAME>.
  • --threads - same as YCSB_THREADS from workload, but with higher priority (overwrites the value from the config).
  • --de-threads - same as YCSB_THREADS_DE from workload, but with higher priority (overwrites the value from the config).
  • --ycsb-hosts - same as YCSB_HOSTS_COUNT, but with higher priority (overwrites the value from the config).

By default scripts use anonymous access to YDB. If you want to use a token, then export YDB_ACCESS_TOKEN_CREDENTIALS (be aware, that currently this token will be printed in the logs and will be visible in process list!).

CockroachDB

Just like with YDB, let's configure the config file cockroach.rc:

  • TARGET - one of the hosts where CockroachDB is running.
  • YCSB_HOSTS - list of hosts on which YCSB will be run.
  • YCSB_HOSTS_COUNT - if you want to limit the number of YCSB_HOSTS without changing the list .
  • COCKROACH_PATH - path to the folder with CockroachDB on YCSB_HOSTS.
  • COCKROACH_TAR_PATH - if COCKROACH_PATH is not present, the archive at this path will be unpacked in COCKROACH_DEPLOY_PATH.
  • HA_PROXY_HOST - one of the hosts where haproxy is running.
  • COCKROACH_INIT_SLEEP_TIME_MINUTES - sometimes export fails with CLI error, but continues in cockroach, so we continue to wait.

After configuring cockroach.rc and workload.rc (more about it above), you can start YCSB:

cd <PATH_TO_BENCHHELPERS>/ycsb
./run_workloads.sh --type cockroach --log-dir <PATH_TO_LOG_DIR> configs/workload.rc configs/cockroach.rc

You can read about additional parameters for run_workloads.sh in YDB.

YugabyteDB

Let's configure the config file yugabyte.rc:

  • TARGET - one of the hosts where YugabyteDB is running.
  • YCSB_HOSTS - list of hosts on which YCSB will be run.
  • YCSB_HOSTS_COUNT - if you want to limit the number of YCSB_HOSTS without changing the list .
  • YU_YCSB_PATH - path to the folder with YSCB of YugabyteDB on YCSB_HOSTS.
  • YU_YCSB_TAR_PATH - if YU_YCSB_PATH is not present, the archive at this path will be unpacked in YU_YCSB_DEPLOY_PATH.
  • YU_PATH - path to the folder with YugabyteDB on TARGET.

After configuring yugabyte.rc and workload.rc (more about it above), you can start YCSB on YCQL:

cd <PATH_TO_BENCHHELPERS>/ycsb
./run_workloads.sh --type yugabyte --log-dir <PATH_TO_LOG_DIR> configs/workload.rc configs/yugabyte.rc

You can read about additional parameters for run_workloads.sh in YDB.

If you want to perform a benchmark on YSQL, then change the --type parameter to yugabyteSQL.