Test Result

This section describes performance metrics that are tested using different data and workload models in different scenarios. Only performance data of instances with small- and medium specifications is presented. To use higher specifications, you can scale out or up the instances.

Table 1 describes the test data used when the data volume is less than memory.
Table 2 describes the test data used when the data volume is greater than memory.

**Table 1** Test data
Instance Specifications	Test Model	Workload Model	QPS	Average Latency (ms)	p99 Latency (ms)
4 vCPUs x 3 nodes	value_length = 100 bytes clients = 90	100% write	125590	0.66	1.85
	value_length = 100 bytes clients = 105	100% read	139741	0.62	1.51
	value_length = 100 bytes clients = 90	50% read and 50% write	125620	Read: 0.56	Read: 1.32
	value_length = 100 bytes clients = 90	50% read and 50% write	125620	Write: 0.55	Write: 1.30
8 vCPUs x 3 nodes	value_length = 100 bytes clients = 128	100% write	216392	0.62	1.92
	value_length = 100 bytes clients = 128	100% read	202970	0.62	1.89
	value_length = 100 bytes clients = 128	50% read and 50% write	212052	Read: 0.63	Read: 1.94
	value_length = 100 bytes clients = 128	50% read and 50% write	212052	Write: 0.63	Write: 1.92

**Table 2** Test data
Instance Specifications	Test Model	Workload Model	QPS	Average Latency (ms)	p99 Latency (ms)
4 vCPUs x 3 nodes	value_length = 100 bytes clients = 75	100% write	123942	0.62	1.30
	value_length = 100 bytes clients = 96	100% read	125351	0.63	1.54
	value_length = 100 bytes clients = 96	50% read and 50% write	122485	Read: 0.64	Read: 1.65
	value_length = 100 bytes clients = 96	50% read and 50% write	122485	Write: 0.64	Write: 1.61
8 vCPUs x 3 nodes	value_length = 100 bytes clients = 120	100% write	196596	0.62	2.02
	value_length = 100 bytes clients = 120	100% read	187716	0.62	1.90
	value_length = 100 bytes clients = 120	50% read and 50% write	197097	Read: 0.62	Read: 1.94
	value_length = 100 bytes clients = 120	50% read and 50% write	197097	Write: 0.62	Write: 1.94