Skytable's benchmark tool sky-bench
is a fairly advanced load-testing tool that can be used to analyze the performance of Skytable installations, and its performance on a certain network with support for full spectrum latency analysis. One of the primary goals of this tool is to provide real-world like workloads and hence as part of the throughput report we output raw throughput and the parsing and validation overhead that you would incur in the real-world when you use Skytable in your applications.
The core of this utility is relatively simple: a thread-pool is created on top of which multiple client "tasks" execute queries. For each query, multiple statistics are computed including:
A workload based approach is used for benchmarking. The currently supported workload is uniform_std_v1
(see the benchmark output for more details on what is executed). Overall, multiple unique rows (1,000,000 unique rows by default) are created with INSERT
, manipulated with an UPDATE
, fetched with a SELECT
and finally deleted with a DELETE
. Hence, for each unique row 4 queries are executed; this means a total of 4,000,000 queries are executed by default.
The workload can be selected using --workload {workload_name}
.
Previously an engine-based approach was used which is now deprecated due to lack of extensibility and other limitations with comprehensive execution analysis. The engines available are rookie
and fury
. The engine can be selected using --engine {engine_name}
.
Please note that we do not recommend using these options and recommend you to use workloads instead.
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。