Views:

INTRODUCTION

Based on the xES provided statistics of agent traffic, Service Gateway performed the performance testing to get the sizing integration with xES.

As the sizing is based on the details of traffics, it may be different with the customer environment. This article can only be used as a reference.

Test Scenarios

Table 1: Regular Agent Network Traffic for testing:

 V1 Agents TypePackage SizeFrequency of polling (API request)
1XBC2.7 MB
  • Polling Next Command: 10 mins
  • Force Polling Next Command: 24 hrs
2WSC1.7 MB
  • Init notify setting: 24 hrs
  • net_module_event: 10 mins
  • AWS SDK ping IoT: 1 min

Table 2: concurrent connection for product request

APIXBC and WSC agents Size
Concurrent connections(agents)10000
Concurrent connections(agents)20000
Concurrent connections(agents)30000

Test Results

Table 2-2 Test Results

PlatformESXI 6.5ESXI 6.5ESXI 6.5ESXI 6.5
Hardware model8vCPU/12G Mem/500G8vCPU/12G Mem/500G8vCPU/12G Mem/500G8vCPU/12G Mem/500G
ProtocolHTTPSHTTPSHTTPSHTTPS
Running statusrunning without loadrunning with loadrunning with loadrunning with load
Client count0100002000030000
Concurrent connection010000 persistent IOT20000 persistent IOT30000 persistent IOT
TPS-20 API request35 API request50 API request
Avg. CPU Usage4.96%6.98%8.13%8.89%
Avg. Mem Usage27.35%33.74%43.90%58.30%
Avg. Disk IO Usage0.91%1.16%1.34%1.42%
Avg. Network Traffic----

TPS: means transaction per second, T(transaction) means API request from a XBC or WSC agent.

Conclusion

One Service Gateway appliance can support at most 30K xES Agents.

By the increase of concurrent connections, SG uses more Mem and CPU, and there is no obvious increase in the usage for the Disk IO, all the resource usage did not exceed 80%.

From the test, the persistent connections mainly cost Mem resource.