Skip to main content

SwiftStack Swift Benchmarking Suite

Project description

A benchmarking suite for the OpenStack Swift object storage system.

The ssbench suite can run benchmark “scenarios” against an OpenStack Swift cluster, saving statistics about the run to a file. It can then generate a report from the saved statstics. By default, a report will be generated to STDOUT immediately following a benchmark run in addition to saving the results to a file.

Coordination between the ssbench-master and one or more ssbench-worker processes is managed through a Beanstalkd queue.

Scenarios

A “scenario” (sometimes called a “CRUD scenario”) is a JSON file defining a benchmark run. Specifically, it defines:

  • A name for the scenario (an arbitrary string)

  • A set of “object size” classes. Each class has a name, a minimum object size and a maximum object size. Objects used within an object size class will have a size (in bytes) chosen at random uniformly between the minimum and maximum sizes.

  • A count of initial files per size class. Each size class can have zero or more objects uploaded prior to the benchmark run itself. The proportion of initial files also defines the probability distribution of object sizes during the benchmark run itself.

  • A count of operations to perform during the benchmark run. An operation is either a CREATE, READ, UPDATE, or DELETE of an object.

  • A “CRUD profile” which determines the distribution of each kind of operation. For instance, [3, 4, 2, 2] would mean 27% CREATE, 36% READ, 18% UPDATE, and 18% DELETE.

  • A user_count which determines the maxiumum client concurrency during the benchmark run. The user is responsible for ensuring there are enough workers running to support the scenario’s defined user_count. (Each ssbench-worker process uses eventlet to achive very efficeint concurrency for the benchmark client requests.)

ssbench comes with a few canned scenarios, but users are encouraged to experiment and define their own.

Here is an example JSON scenario file:

{
  "name": "Small test scenario",
  "sizes": [{
    "name": "tiny",
    "size_min": 4096,
    "size_max": 65536
  }, {
    "name": "small",
    "size_min": 100000,
    "size_max": 200000
  }],
  "initial_files": {
    "tiny": 100,
    "small": 10
  },
  "operation_count": 500,
  "crud_profile": [3, 4, 2, 2],
  "user_count": 7
}

Installation

Install this module (ssbench) via pip. You will also need Beanstalkd and an OpenStack Swift cluster to benchmark.

Usage

$ ssbench-worker --help
usage: ssbench-worker [-h] [--qhost QHOST] [--qport QPORT] [--retries RETRIES]
                      [--concurrency CONCURRENCY]
                      worker_id

Benchmark your Swift installation

positional arguments:
  worker_id             An integer ID number; must be unique among all workers

optional arguments:
  -h, --help            show this help message and exit
  --qhost QHOST         beanstalkd host (def: localhost)
  --qport QPORT         beanstalkd port (def: 11300)
  --retries RETRIES     Maximum number of times to retry a job.
  --concurrency CONCURRENCY
                        Number of concurrent connections for this worker
$ ssbench-master -h
usage: ssbench-master [-h] [--qhost QHOST] [--qport QPORT]
                      {run-scenario,report-scenario} ...

Benchmark your Swift installation

positional arguments:
  {run-scenario,report-scenario}
    run-scenario        Run CRUD scenario, saving statistics
    report-scenario     Generate a report from saved scenario statistics

optional arguments:
  -h, --help            show this help message and exit
  --qhost QHOST         beanstalkd host (default: localhost)
  --qport QPORT         beanstalkd port (default: 11300)
$ ssbench-master run-scenario -h
usage: ssbench-master run-scenario [-h] -A AUTH_URL -K KEY -U USER -f
                                   SCENARIO_FILE [-s STATS_FILE] [-r]

optional arguments:
  -h, --help            show this help message and exit
  -A AUTH_URL, --auth-url AUTH_URL
  -K KEY, --key KEY
  -U USER, --user USER
  -f SCENARIO_FILE, --scenario-file SCENARIO_FILE
  -s STATS_FILE, --stats-file STATS_FILE
                        File into which benchmarking statistics will be saved
                        (default: /tmp/ssbench-results/<scenario_name>.stat)
  -r, --no-default-report
                        Suppress the default immediate generation of a
                        benchmark report to STDOUT after saving stats-file
                        (default: False)
$ ssbench-master report-scenario -h
usage: ssbench-master report-scenario [-h] -s STATS_FILE [-f REPORT_FILE]
                                      [-r RPS_HISTOGRAM]

optional arguments:
  -h, --help            show this help message and exit
  -s STATS_FILE, --stats-file STATS_FILE
                        An existing stats file from a previous --run-scenario
                        invocation (default: None)
  -f REPORT_FILE, --report-file REPORT_FILE
                        The file to which the report should be written (def:
                        STDOUT) (default: <open file '<stdout>', mode 'w' at
                        0x1002511e0>)
  -r RPS_HISTOGRAM, --rps-histogram RPS_HISTOGRAM
                        Also write a CSV file with requests completed per
                        second histogram data (default: None)

Example Run

First make sure beanstalkd is running.

$ beanstalkd -l 127.0.0.1 &

Then, start one or more ssbench-worker processes.

$ ssbench-worker 1 &
$ ssbench-worker 2 &

Finally, run one ssbench-master process which will manage and coordinate the benchmark run.

$ ssbench-master run-scenario -A http://192.168.22.100/auth/v1.0 -U dev:admin -K admin -f very_small.scenario
INFO:root:Starting scenario run for u'Small test scenario'
INFO:root:Creating containers (ssbench_*) with concurrency 10...
INFO:root:Initializing cluster with stock data (up to 7 concurrent workers)
INFO:root:Starting benchmark run (up to 7 concurrent workers)
INFO:root:Deleting population objects from cluster
INFO:root:Calculating statistics for 500 result items...

Small test scenario
  C   R   U   D     Worker count:   2   Concurrency:   7
% 27  36  18  18

TOTAL
       Count:   500  Average requests per second:  45.5
                           min      max     avg     std_dev   median
       First-byte latency:  0.01 -   0.44    0.11  (  0.09)    0.07  (  all obj sizes)
       Last-byte  latency:  0.01 -   0.44    0.11  (  0.09)    0.07  (  all obj sizes)
       First-byte latency:  0.01 -   0.44    0.11  (  0.09)    0.07  (tiny objs)
       Last-byte  latency:  0.01 -   0.44    0.11  (  0.09)    0.07  (tiny objs)
       First-byte latency:  0.01 -   0.38    0.13  (  0.10)    0.09  (small objs)
       Last-byte  latency:  0.01 -   0.38    0.13  (  0.10)    0.09  (small objs)

CREATE
       Count:   133  Average requests per second:  12.3
                           min      max     avg     std_dev   median
       First-byte latency:  0.03 -   0.44    0.16  (  0.10)    0.13  (  all obj sizes)
       Last-byte  latency:  0.03 -   0.44    0.16  (  0.10)    0.14  (  all obj sizes)
       First-byte latency:  0.03 -   0.44    0.16  (  0.10)    0.13  (tiny objs)
       Last-byte  latency:  0.03 -   0.44    0.16  (  0.10)    0.13  (tiny objs)
       First-byte latency:  0.08 -   0.38    0.20  (  0.10)    0.23  (small objs)
       Last-byte  latency:  0.08 -   0.38    0.21  (  0.10)    0.23  (small objs)

READ
       Count:   176  Average requests per second:  16.2
                           min      max     avg     std_dev   median
       First-byte latency:  0.01 -   0.16    0.04  (  0.03)    0.03  (  all obj sizes)
       Last-byte  latency:  0.01 -   0.16    0.04  (  0.03)    0.03  (  all obj sizes)
       First-byte latency:  0.01 -   0.16    0.04  (  0.03)    0.03  (tiny objs)
       Last-byte  latency:  0.01 -   0.16    0.04  (  0.03)    0.03  (tiny objs)
       First-byte latency:  0.01 -   0.08    0.04  (  0.02)    0.04  (small objs)
       Last-byte  latency:  0.01 -   0.08    0.04  (  0.02)    0.04  (small objs)

UPDATE
       Count:   100  Average requests per second:   9.2
                           min      max     avg     std_dev   median
       First-byte latency:  0.03 -   0.36    0.15  (  0.08)    0.13  (  all obj sizes)
       Last-byte  latency:  0.03 -   0.36    0.15  (  0.08)    0.13  (  all obj sizes)
       First-byte latency:  0.03 -   0.36    0.14  (  0.08)    0.13  (tiny objs)
       Last-byte  latency:  0.03 -   0.36    0.14  (  0.08)    0.13  (tiny objs)
       First-byte latency:  0.06 -   0.33    0.20  (  0.09)    0.21  (small objs)
       Last-byte  latency:  0.08 -   0.33    0.20  (  0.08)    0.21  (small objs)

DELETE
       Count:    91  Average requests per second:   8.3
                           min      max     avg     std_dev   median
       First-byte latency:  0.02 -   0.33    0.11  (  0.08)    0.12  (  all obj sizes)
       Last-byte  latency:  0.02 -   0.33    0.11  (  0.08)    0.12  (  all obj sizes)
       First-byte latency:  0.02 -   0.33    0.12  (  0.08)    0.12  (tiny objs)
       Last-byte  latency:  0.02 -   0.33    0.12  (  0.08)    0.12  (tiny objs)
       First-byte latency:  0.03 -   0.14    0.07  (  0.04)    0.04  (small objs)
       Last-byte  latency:  0.03 -   0.14    0.07  (  0.04)    0.04  (small objs)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

ssbench-0.0.6.tar.gz (35.3 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page