We are trying to benchmark HTTP Throughput with netty.io based HTTP server. Both the client and server systems have the following HW and SW resources: Lenovo ThinkSystem SR630, 250GB RAM, Intel Xeon Gold 5120 CPU @2.20GHz CPU model, 28 CPU Cores, Solarflare X2541 Network Adapter with 100GbE link speed, RHEL 7.5. The netty.io HTTP server code was taken from the https://netty.io/4.1/xref/io/netty/example/http/snoop/package-summary.html and the wrk2 https://github.com/giltene/wrk2 is used as load generating tool. The following are the results for 256 byte request size with 1000 wrk client connections (wrk -t<cores> -c1000 -d30s -R 10M -L http://Server_IP:8080)
Cores Kernel Requests/sec 1 Core 53,976 4 Cores 201,272 8 Cores 205,635 12 Cores 197,908 16 Cores 197,730 20 Cores 197,252 24 Cores 195,801 However, we came across the https://github.com/TechEmpower/FrameworkBenchmarks/tree/master/frameworks/Java/netty which shows its throughput result for netty on Best plaintext responses per second, Dell R440 Xeon Gold + 10 GbE is 4,648,537 for around the same request size of 256 bytes. We repeated the test with the same netty code provided in the https://github.com/TechEmpower/FrameworkBenchmarks/tree/master/frameworks/Java/netty, but the results are similar to the results table above. We are just wondering what makes such a huge difference. The netty code is more or less similar. Is there anything that stands out? -- You received this message because you are subscribed to the Google Groups "Netty discussions" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/netty/2fc7c292-8021-4688-b8eb-e74f7fce65d8%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
