Sysbench Report

This scenario is executed with Sysbench tool. There is one instance of tool per tester node, each running in N threads.

Throughput

The following chart shows the number of queries, read queries and transactions depending on total thread count.

{{''': title: Throughput axes: x: threads y1: queries per sec y2: read queries per sec y3: transactions per sec chart: line pipeline: - $match: { task: sysbench_oltp } - $group: _id: { threads: { $multiply: [ "$threads", "$host_count" ] } } queries_total_per_sec: { $sum: { $divide: ["$queries_total", "$duration"] }} queries_read_per_sec: { $sum: { $divide: ["$queries_read", "$duration"] }} transactions_per_sec: { $sum: { $divide: ["$transactions", "$duration"] }} - $project: x: "$_id.threads" y1: "$queries_total_per_sec" y2: "$queries_read_per_sec" y3: "$transactions_per_sec"

''' | chart_and_table }}

The following chart shows how DB server CPU consumption depends on number of concurrent threads and throughput.

{{''': title: CPU consumption axes: x: threads y1: queries per sec y2: CPU, % chart: line pipeline: - $match: { task: sysbench_oltp } - $group: _id: { threads: { $multiply: [ "$threads", "$host_count" ] } } queries_total_per_sec: { $sum: { $divide: ["$queries_total", "$duration"] }} mysqld_total: { $avg: "$mysqld_total" } - $project: x: "$_id.threads" y1: "$queries_total_per_sec" y2: { $multiply: [ "$mysqld_total", 100 ] }

''' | chart_and_table }}

The following chart shows how operation latency depends on number of concurrent threads.

{{''': title: Latency axes: x: threads y1: min latency, ms y2: avg latency, ms y3: max latency, ms chart: line pipeline: - $match: { task: sysbench_oltp } - $group: _id: { threads: { $multiply: [ "$threads", "$host_count" ] } } latency_min: { $min: "$latency_min" } latency_avg: { $avg: "$latency_avg" } latency_max: { $max: "$latency_max" } - $project: x: "$_id.threads" y1: "$latency_min" y2: "$latency_avg" y3: "$latency_max"

''' | chart_and_table }}